An automatic online and offline circuit for data transmission

Abstract

The present invention belongs to the technical field of chip development, and specifically discloses an automatic online and offline circuit for data transmission. The circuit connection and working method proposed by the present invention enable the sending node and the receiving node circuits to be connected in a pipelined manner when data is transmitted within the chip. One data transmission configuration can be put online in each clock cycle, which can significantly reduce the delay of caching the ID and its configuration during data transmission within the chip, and effectively improve the efficiency of data transmission within the chip.

Description

Technical Field

[0001] The present invention belongs to the technical field of chip development, and particularly relates to an automatic online and offline circuit for data transmission. Background Art

[0002] In the on-chip system of an NPU chip, a large amount of data transmission is involved. Each data transmission is identified by a unique ID representing the identity, and at the same time, the ID also uniquely represents the configuration information required for data transmission. However, for the on-chip system of an NPU chip, due to the limited cache in the sending node and the receiving node, it is impossible to store all the transmitted IDs and their corresponding configurations. Therefore, after all the IDs and configuration information are generated by the compiler, they are stored in the memory outside the sending node and the receiving node. When a transmission occurs, the ID and its configuration need to be cached into the sending node and the receiving node first, and this process is called online; after the transmission is completed, the storage space occupied by the unnecessary IDs and their configurations is released from the sending node and the receiving node, and this process is called offline.

[0003] Traditionally, the online and offline functions in the NPU on-chip system are implemented by software, which will result in a large delay in the process of loading the ID and its configuration information into the hardware cache, and the size of this delay will directly affect the performance of the NPU on-chip system. Summary of the Invention

[0004] In order to solve at least one of the problems mentioned in the above background art, the present invention proposes an automatic online and offline circuit for data transmission.

[0005] In the present invention, online refers to the process from the online instruction sent by the data generation calculation module to the ID and configuration information being loaded into the configuration information cache module; offline refers to the process from the offline instruction sent by the data generation calculation module to the space occupied by the ID and configuration information being released from the configuration information cache module.

[0006] An automatic online and offline circuit for data transmission specifically includes:

[0007] A data generation and calculation module for generating and consuming data and controlling the data online and offline processes; an online instruction cache module for caching online instructions; a configuration information reading module for reading configuration information required for data transmission; a configuration information read request cache module for caching read requests sent by the configuration information reading module to the memory; a configuration information cache module for caching online and offline configuration information; a configuration information management module for monitoring and recording handshake status and managing data address allocation; an online response cache module for caching successful online responses; an online and offline configuration cache module for caching online and offline instruction configuration information; an on-chip interconnection network module for connecting all sending nodes and receiving nodes; a data address management module for managing data addresses; a register path module for providing a register path; an online instruction counting module for counting the number of online instructions;

[0008] The automatic online and offline circuit for data transmission includes a main line one connection, a main line two connection, a branch line one connection, a branch line two connection, a branch line three connection, and a branch line four connection. Among them,

[0009] The connection method of the main line one is: connecting the data generation and calculation module, the online instruction cache module, the configuration information reading module, the online and offline configuration cache module, the configuration information management module, the data address management module, and the register path module; the connection method of the main line two is: connecting the configuration information management module, the online response cache module, and the data generation and calculation module;

[0010] The connection method of the branch line one is: connecting the data generation and calculation module with the online instruction counting module; the connection method of the branch line two is: first connecting the configuration information reading module, the configuration information read request cache module, and the memory, and then connecting the memory and the configuration information reading module; the connection method of the branch line three is: connecting the configuration information management module with the on-chip interconnection network module; the connection method of the branch line four is: connecting the configuration information management module with the configuration information cache module.

[0011] A working method of an automatic online and offline circuit for data transmission includes a receiving node working process and a sending node working process. Among them,

[0012] The receiving node working process is:

[0013] First, the data generation and calculation module issues an online instruction, and this online instruction is cached in the online instruction cache module. The specific online instruction is: prefix ID + ready, and the specific offline instruction is: prefix ID -.

[0014] Then, after the receiving node completes the online process, it will notify the sending node through the on-chip interconnection network module;

[0015] Next, when the sending node is online and receives the notification signal from the receiving node, it sends data to the receiving node.

[0016] Furthermore, the receiving node receives the data sent by the sending node and writes the data into local storage.

[0017] Finally, when the receiving node receives the offline command, it releases the space occupied by the ID and configuration information from the configuration information cache module.

[0018] The working process of the sending node is as follows:

[0019] First, a data generation calculation module issues an online instruction, which is cached in the online instruction cache module. The specific online instruction is: prefix ID + ready, and the specific offline instruction is: prefix ID -.

[0020] Then, after the sending node completes going online, it waits for the ready signal from the receiving node.

[0021] Next, when the sending node receives the ready signal from the receiving node, it starts to read data from local storage.

[0022] Furthermore, the read data is sent to the receiving node.

[0023] Finally, when the sending node receives the offline command, it releases the space occupied by the ID and configuration information from the configuration information cache module.

[0024] Traditionally, the online and offline functions in the NPU on-chip system are implemented by software, which results in a large delay in the process of loading the ID and its configuration information into the hardware cache. The size of this delay will directly affect the performance of the NPU on-chip system. The circuit connection and its corresponding working method designed in the present invention enable the sending node and the receiving node circuits to be connected in a pipelined manner when data is transmitted within the chip. One data transmission configuration can be put online per clock cycle, which can significantly reduce the delay of caching the ID and its configuration during data transmission within the chip and effectively improve the efficiency of data transmission within the chip.

[0025] The present invention proposes an automatic online and offline circuit for data transmission. Compared with the existing technologies, it has the following beneficial effects:

[0026] The circuit connection and its working method proposed in the present invention enable the sending node and the receiving node circuits to be connected in a pipelined manner when data is transmitted within the chip. One data transmission configuration can be put online per clock cycle, which can significantly reduce the delay of caching the ID and its configuration during data transmission within the chip and effectively improve the efficiency of data transmission within the chip. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is the process schematic diagram of the present invention;

[0028] Figure 2 is the schematic diagram from the timing perspective of the receiving node in the embodiment of the present invention;

[0029] Figure 3 is the schematic diagram from the timing perspective of the sending node in the embodiment of the present invention. Detailed implementation manners

[0030] In order to make the objectives and features of the present invention more obvious and understandable, the following describes the technical solution in detail through embodiments and in conjunction with the drawings.

[0031] In the present invention, going online refers to the process from the data generation and calculation module sending a going-online instruction to the ID and configuration information being loaded into the configuration information cache module; going offline refers to the process from the data generation and calculation module sending a going-offline instruction to releasing the space occupied by the ID and configuration information from the configuration information cache module.

[0032] As Figure 1 shown, an automatic online and offline circuit for data transmission specifically includes:

[0033] A data generation and calculation module for generating and consuming data and controlling the data online and offline processes; an online instruction cache module for caching online instructions; a configuration information reading module for reading configuration information required for data transmission; a configuration information read request cache module for caching read requests sent by the configuration information reading module to the memory; a configuration information cache module for caching online and offline configuration information; a configuration information management module for monitoring and recording handshake status and managing data address allocation; an online response cache module for caching online success responses; an online and offline configuration cache module for caching online and offline instruction configuration information; an on-chip interconnection network module for connecting all sending nodes and receiving nodes; a data address management module for managing data addresses; a register path module for providing a register path; an online instruction counting module for counting the number of online instructions;

[0034] The automatic online and offline circuit for data transmission includes a first trunk connection, a second trunk connection, a first branch connection, a second branch connection, a third branch connection, and a fourth branch connection. Among them,

[0035] The connection method of the first trunk is: connecting the data generation and calculation module, the online instruction cache module, the configuration information reading module, the online and offline configuration cache module, the configuration information management module, the data address management module, and the register path module; the connection method of the second trunk is: connecting the configuration information management module, the online response cache module, and the data generation and calculation module.

[0036] The connection method of branch line 1 is: connecting the data generation and calculation module with the online instruction counting module; the connection method of branch line 2 is: first connecting the configuration information reading module, the configuration information read request cache module and the memory, and then connecting the memory and the configuration information reading module; the connection method of branch line 3 is: connecting the configuration information management module with the on-chip interconnection network module; the connection method of branch line 4 is: connecting the configuration information management module with the configuration information cache module.

[0037] A working method of an automatic online and offline circuit for data transmission includes a receiving node working process and a sending node working process, wherein,

[0038] The receiving node working process is:

[0039] First, a data generation and calculation module issues an online instruction, and this online instruction is cached in the online instruction cache module. The specific online instruction is: prefix ID + ready, and the specific offline instruction is: prefix ID -.

[0040] Then, after the receiving node completes the online process, it will notify the sending node through the on-chip interconnection network module.

[0041] Next, when the sending node is already online and receives the notification signal from the receiving node, it sends data to the receiving node.

[0042] Further, the receiving node receives the data sent by the sending node and writes this data into local storage.

[0043] Finally, when the receiving node receives the offline command, it releases the space occupied by the ID and the configuration information from the configuration information cache module.

[0044] The timing perspective of the receiving node is as Figure 2 shown.

[0045] The sending node working process is:

[0046] First, a data generation and calculation module issues an online instruction, and this online instruction is cached in the online instruction cache module. The specific online instruction is: prefix ID + ready, and the specific offline instruction is: prefix ID -.

[0047] Then, after the sending node completes the online process, it will wait for the ready signal from the receiving node.

[0048] Next, when the sending node receives the ready signal from the receiving node, it starts to read data from local storage.

[0049] Further, the read data is sent to the receiving node.

[0050] Finally, when the sending node receives the offline command, it releases the space occupied by the ID and the configuration information from the configuration information cache module.

[0051] The timing perspective of the sending node is as Figure 3 shown.

[0052] Thus, according to the method disclosed in the present invention, one working process of the present invention has been completed.

[0053] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments. And the term "including" or "comprising" or any other variant thereof in this article is intended to cover a non-exclusive inclusion, so that a process, apparatus, article or method including a series of elements not only includes those elements but also includes other elements not expressly listed, or also includes elements inherent to such process, apparatus, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, apparatus, article or method including that element.

[0054] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above and includes several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0055] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An automatic online and offline circuit for data transmission, characterized in that, including: a data generation and calculation module for generating and consuming data and controlling the data online and offline processes; an online instruction cache module for caching online instructions; a configuration information reading module for reading configuration information required for data transmission; a configuration information read request cache module for caching read requests sent by the configuration information reading module to the memory; a configuration information cache module for caching online and offline configuration information; a configuration information management module for monitoring and recording handshake status and managing data address allocation; an online response cache module for caching successful online responses; an online and offline configuration cache module for caching online and offline instruction configuration information; an on-chip interconnection network module for connecting all sending nodes and receiving nodes; a data address management module for managing data addresses; a register path module for providing a register path; an online instruction counting module for counting the number of online instructions; The online process refers to the process from the data generation and calculation module sending an online instruction to the ID and configuration information being loaded into the configuration information cache module. The offline process refers to the process from the data generation and calculation module sending an offline instruction to releasing the space occupied by the ID and configuration information from the configuration information cache module.

2. The automatic online and offline circuit for data transmission according to claim 1, wherein The automatic online and offline circuit includes a main line connection and branch line connections. Among them, the main line connection includes a main line one connection and a main line two connection; the branch line connections include a branch line one connection, a branch line two connection, a branch line three connection, and a branch line four connection.

3. The automatic online and offline circuit for data transmission according to claim 2, characterized in that The main line one connection method is: connecting the data generation and calculation module, the online instruction cache module, the configuration information reading module, the online and offline configuration cache module, the configuration information management module, the data address management module, and the register path module; The main line two connection method is: connecting the configuration information management module, the online response cache module, and the data generation and calculation module; The branch line one connection method is: connecting the data generation and calculation module to the online instruction counting module; The branch line two connection method is: first connecting the configuration information reading module, the configuration information read request cache module, and the memory, and then connecting the memory and the configuration information reading module; The branch line three connection method is: connecting the configuration information management module to the on-chip interconnection network module; The branch line four connection method is: connecting the configuration information management module to the configuration information cache module.

4. An automatic online / offline circuit for data transmission according to claim 1, characterized in that The automatic online and offline circuit includes a receiving node working process and a sending node working process.

5. The automatic online and offline circuit for data transmission according to claim 4, characterized in that The receiving node working process includes the steps: Step A1, the data generation and calculation module sends an online instruction, and this online instruction is cached in the online instruction cache module. The specific online instruction is: prefix ID + ready, and the specific offline instruction is: prefix ID -; Step A2, after the receiving node completes the online process, it will notify the sending node through the on-chip interconnection network module; Step A3, when the sending node is already online and receives the notification signal from the receiving node, it will send data to the receiving node; Step A4, the receiving node receives the data sent by the sending node and writes the data into local storage; Step A5, when the receiving node receives the offline command, it releases the space occupied by the ID and the configuration information from the configuration information cache module.

6. The automatic online and offline circuit for data transmission according to claim 4, characterized in that, The working process of the sending node includes the steps: Step B1, a line-up instruction is issued by the data generation and calculation module, and the line-up instruction is cached in the line-up instruction cache module. The specific line-up instruction is: prefix ID + ready, and the offline instruction is specifically: prefix ID-. Step B2, after the sending node completes the line-up, it waits for the ready signal from the receiving node; Step B3, when the sending node receives the ready signal from the receiving node, it starts to read data from local storage; Step B4, sends the read data to the receiving node; Step B5, when the sending node receives the offline command, it releases the space occupied by the ID and the configuration information from the configuration information cache module.

Images