Data transmission method and related device

By introducing a waiting confirmation signal mechanism in UCIE communication, data backpressure of data packet transmission between core particles is realized, packet overflow and loss is avoided, transmission efficiency and stability are improved, and data transmission efficiency is solved in the existing technology is low data transmission efficiency The problem.

CN118981444APending Publication Date: 2024-11-19PHYTIUM TECH CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411008955.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

During the UCIE-based inter-core communication process, the problem of receiving data overflow or receiving data loss may occur, and the lack of an effective data backpressure mechanism leads to a reduction in data transmission efficiency.

Method used

By introducing a waiting confirmation signal mechanism in the adaptation layer, we decide whether to discard the data packet based on the storage space of the data cache queue, and actively initiate data backpressure at the receiving end to avoid overflow or loss of data packets, and at the same time stop the relevant timer to avoid Reset or retrain the link by mistake.

Benefits of technology

It effectively avoids packet overflow and loss, improves data transmission efficiency, reduces invalid link reset and retraining, and improves the stability and efficiency of inter-core communication.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981444A_ABST
    Figure CN118981444A_ABST
Patent Text Reader

Abstract

According to the data transmission method provided by the embodiment of the invention, a data back pressure mechanism in a data receiving direction is realized based on an adaptation layer of a UCIE module, support is provided for a data back pressure function when a user transmits a customized data stream, and the phenomenon that the data back pressure function is influenced when the user transmits the customized data stream is avoided. And the problem of data loss caused by data overflow of a data cache region of a receiving end is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present specification relates to the field of communication technology, specifically, to chip-to-chip communication technology in the field of communication technology, and more specifically, to a data transmission method and related devices. Background Art

[0002] UCIE (Universal Chiplet Interconnect Express) is an open chiplet interconnect protocol that aims to provide a universal interconnect framework between chipsets. In chipset technology, multiple chiplets are combined in a package to form a complete system or processor. These chiplets can include any type of chip such as CPU (Central Processing Unit) core, GPU (Graphics Processing Unit) core, memory controller, I / O (Input / Output) interface, etc. The purpose of UCIE is to provide a standardized interface and interconnection protocol for these different chips to facilitate communication between them.

[0003] Currently, in the process of inter-chiplet communication based on UCIE, problems of received data overflow or received data loss may occur. Summary of the invention

[0004] The embodiments of this specification provide a data transmission method and related devices to implement the function of receiving data back pressure during the communication between core particles, thereby avoiding the problems of received data loss and received data overflow.

[0005] To achieve the above technical objectives, the embodiments of this specification provide the following technical solutions:

[0006] An embodiment of the present specification provides a data transmission method, which is applied to a first coregrain in a system on chip, wherein the first coregrain includes a first protocol layer and a first universal coregrain interconnect technology UCIE module, wherein the first protocol layer includes a data cache queue, wherein the data cache queue is used to cache data packets, wherein the first UCIE module includes a first adaptation layer, wherein the first adaptation layer includes a first timer, wherein the system on chip further includes a second coregrain, wherein the second coregrain includes a second UCIE module, wherein the second UCIE module includes a second adaptation layer, wherein the second adaptation layer includes a retransmission buffer and a second timer, wherein the data transmission method includes:

[0007] In response to the current data packet sent by the second core particle, when the value of the first flag signal is the first target value, the first adaptation layer discards the current data packet, returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer, stops and resets the first timer, and the waiting confirmation signal is used to instruct the second adaptation layer to control the retransmission buffer to resend the current data packet according to the received sequence value, and stops and resets the second timer;

[0008] The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

[0009] In a second aspect, an embodiment of the present specification provides a data transmission method, which is applied to a second coregrain in a system on chip, wherein the second coregrain includes a second UCIE module, the second UCIE module includes a second adaptation layer, the second adaptation layer includes a retransmission buffer and a second timer, and the system on chip further includes: a first coregrain, the first coregrain includes a first protocol layer and a first UCIE module, the first protocol layer includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module includes a first adaptation layer, and the first adaptation layer includes a first timer; the data transmission method includes:

[0010] The second adaptation layer sends a current data packet to the first adaptation layer, where the current data packet is used to indicate that the first adaptation layer, in response to the current data packet sent by the second core particle, discards the current data packet when the value of the first flag signal is the first target value, returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer, and stops and resets the first timer;

[0011] The second adaptation layer controls the retransmission buffer to resend the current data packet according to the received sequence value in response to the waiting confirmation signal, and stops and resets the second timer;

[0012] The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

[0013] In a third aspect, an embodiment of the present specification further provides a system on chip, including: a first core particle and a second core particle; the first core particle includes a first protocol layer and a first UCIE module, the first protocol layer includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module includes a first adaptation layer, the first adaptation layer includes a first timer, the second core particle includes a second UCIE module, the second UCIE module includes a second adaptation layer, the second adaptation layer includes a retransmission buffer and a second timer; wherein,

[0014] The second adaptation layer is configured to: send the current data packet to the first adaptation layer, in response to the waiting confirmation signal, control the retransmission buffer to resend the current data packet according to the received sequence value carried by the waiting confirmation signal, and stop and reset the second timer;

[0015] The first adaptation layer is configured to: in response to a current data packet sent by the second core particle, when the value of the first flag signal is the first target value, discard the current data packet, return a waiting confirmation signal carrying a received sequence value to the second adaptation layer, and stop and reset the first timer;

[0016] The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

[0017] In a fourth aspect, an embodiment of the present specification further provides a computing device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the data transmission method as described above when executing the computer program.

[0018] In a fifth aspect, an embodiment of the present specification further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the data transmission method as described above is implemented.

[0019] In a sixth aspect, an embodiment of the present specification provides a computer program product or a computer program, wherein the computer program product includes a computer program, and the computer program is stored in a computer-readable storage medium; the processor of the computer device reads the computer program from the computer-readable storage medium, and the processor implements the steps of the above-mentioned data transmission method when executing the computer program.

[0020] It can be seen from the above technical solution that in the data transmission method provided in the embodiment of this specification, after the first adaptation layer receives the current data packet sent by the second core particle, it determines whether to perform data back pressure according to whether the value of the first flag signal is the first target value, so as to avoid receiving a large number of data packets to the first protocol layer and causing data packet overflow or loss. Specifically, when the value of the first flag signal is the first target value, it indicates that the storage space of the data cache queue in the first protocol layer is full. At this time, the number of data packets waiting to be processed by the first protocol layer is large. If data packets are continuously received, the data packets received by the first protocol layer may overflow or be lost. Therefore, when the value of the first flag signal is the first target value, the first adaptation layer discards the current data packet received, stops and resets the first timer, and avoids the erroneous sending of the response signal; at the same time, the first adaptation layer returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer to instruct the second adaptation layer to control the retransmission buffer to resend the current data packet according to the received sequence value, so that when the first protocol layer can process the data packet normally, the first adaptation layer can receive the data packet in time and transmit it to the first protocol layer; at the same time, the waiting confirmation signal also instructs the second adaptation layer to stop and reset the second timer, so that the second adaptation layer clearly knows that there is no data packet error or link transmission error when the waiting confirmation signal is received, avoiding the reset or retraining of the data link when the second timer frequently times out the sending time. In this way, the data transmission method realizes data back pressure actively initiated by the receiving end (the first core particle) when data packets are transmitted between core particles, and at the same time avoids the incorrect resetting or retraining of the data link when the sending end (the second core particle) frequently sends the same data packet under the data back pressure state, thereby avoiding the reduction of data transmission efficiency caused by invalid resetting and invalid retraining of the data link in this case, which is beneficial to improving the data transmission efficiency of the data transmission method. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0022] Figure 1 This is a schematic diagram of the application scenarios of UCIE technology;

[0023] Figure 2 A schematic diagram of a data packet format of a 256B flit mode (256-byte data unit mode) after delay optimization;

[0024] Figure 3This is a schematic diagram of data back pressure at the sending end;

[0025] Figure 4 Schematic diagram of data received by the receiving end;

[0026] Figure 5 It is a schematic diagram of the protocol layer, adaptation layer and physical layer in the receiving end;

[0027] Figure 6 A schematic diagram of a packet header of a data packet;

[0028] Figure 7 It is a schematic diagram of inter-chiplet communication based on UCIE technology in the related art;

[0029] Figure 8 A flowchart of a data transmission method provided for one embodiment of this specification;

[0030] Fig. 9 A feasible flow chart of a data transmission method provided for one embodiment of this specification;

[0031] Fig.10 A flowchart of an ACK mechanism provided for one implementation of this specification;

[0032] Fig.11 A flowchart of a NAK mechanism provided for one embodiment of this specification;

[0033] Fig.12 A flowchart of a WACK mechanism provided for one embodiment of this specification;

[0034] Fig.13 A flowchart of another data transmission method provided for one embodiment of this specification;

[0035] Fig.14 A schematic diagram of the structure of a system on a chip provided for one embodiment of the present specification;

[0036] Fig.15 A schematic diagram of the structure of a computing device provided for one embodiment of the present specification. DETAILED DESCRIPTION

[0037] Unless otherwise defined, the technical terms or scientific terms used in the embodiments of this specification shall have the common meanings understood by persons with ordinary skills in the field to which this specification belongs. The words "first", "second" and similar words used in the embodiments of this specification do not indicate any order, quantity or importance, but are only used to avoid confusion of constituent elements.

[0038] Unless the context requires otherwise, throughout the specification, "plurality" means "at least two", and "including" is interpreted as an open, inclusive meaning, that is, "including, but not limited to". In the description of the specification, the terms "one embodiment", "some embodiments", "exemplary embodiments", "examples", "specific examples" or "some examples" are intended to indicate that a particular feature, structure, material or characteristic associated with the embodiment or example is included in at least one embodiment or example of the specification. The schematic representation of the above terms does not necessarily refer to the same embodiment or example.

[0039] The technical solutions in the embodiments of this specification will be described below in conjunction with the drawings in the embodiments of this specification. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.

[0040] Overview

[0041] refer to Figure 1 , Figure 1 The basic application scenario of UCIE technology is shown. Two core particles (dieA and dieB) are connected through the UCIE protocol. Each core particle integrates a UCIE module. The protocol layer sends data to the UCIE module or receives data from the UCIE module through the FDI (Flit-Aware Die-to-DieInterface) interface. Inside the UCIE module, the adaptation layer and the physical layer interact with each other through the RDI (Raw Die-to-Die Interface) interface. The two core particles are connected through a physical link. TX indicates the data sending direction and RX indicates the data receiving direction.

[0042] Backpressure is a flow control mechanism that manages packets in a data stream to prevent the sender (TX) from overflowing the receiver's (RX) buffer. When the receiver processes data slower than the sender can send data, the backpressure mechanism allows the receiver to notify the sender in some way to slow down or pause sending until the receiver can catch up. This mechanism prevents data loss and buffer overflow.

[0043] Based on UCIE technology, the two cores can transmit data packets such as CXL (Compute Express Link) and PCIe (Peripheral Component Interconnect Express). When transmitting data packets such as CXL and PCIe, the protocol layer can be used to integrate credit-based flow control in the data stream to achieve the data back pressure function of the receiving core. Figure 2 , Figure 2 The data packet format of 256B flit mode (256-byte data unit mode) after delay optimization is shown. Figure 2 In the figure, 62B of FlitChunk i (i=0~3) (from Protocol Layer) respectively represent the 62-byte data unit blocks transmitted from the protocol layer to the adaptation layer, and the data to be transmitted is placed in these parts. FH B0 / B1 respectively represent Flit Header (data unit header, or packet header) Byte 0 (byte 0) and Byte 1 (byte 1). C0 B0 / B1, C1 B0 / B1 respectively represent CRC0 Byte 0, CRC0 Byte 1, CRC1 Byte 0, and CRC1 Byte 1, i.e., the bytes used for CRC (Cyclic Redundancy Check) verification.

[0044] In the credit-based flow control function, the sending of packets is managed by using "credits", in which the receiver allocates credits to the sender based on its available buffer space. The sender must have enough credits to send packets. Every time a packet is sent, the sender's credits decrease; every time the receiver processes a packet and frees up buffer space, it increases the sender's credits. This mechanism ensures that the sender does not send more data than the receiver can handle.

[0045] However, the "credit" information needs to be integrated into the data unit block (i.e. Flit Chunk), which squeezes the storage space of the data that really needs to be sent. On the one hand, it will increase the complexity of the data packet. On the other hand, the sending and receiving of "credit" will occupy a certain amount of data bandwidth, reducing the bandwidth utilization during the real data transmission process.

[0046] In the mode of transmitting user-defined data streams (such as Streaming Protocol), UCIE provides a data back pressure mechanism in the data sending direction (that is, in the same core, data is sent from the protocol layer to the adaptation layer, which is called the sending direction, or TX direction). Figure 3 and Figure 4 , lp* indicates the signal output by the protocol layer, and pl_* indicates the signal output by the adaptation layer to the protocol layer. Figure 3 It can be found that the adaptation layer can realize data back pressure by pulling down the pl_trdy signal, so that the protocol layer can stop sending new data (that is, when the pl_trdy signal is pulled low, the lp_data signal remains at Dat2, that is, data sending is stopped).

[0047] But in the data receiving direction, refer to Figure 4 Although it supports receiving a variety of data formats, these data formats do not support the back pressure function, that is, the receiving end chip has no suitable mechanism to make the adapter layer of the sending end chip stop sending new data. All data sent from the sending end chip, as long as it is sent to the protocol layer through the receiving end's adaptation layer, must be received by the protocol layer. Figure 3 and Figure 4 In the , *_data indicates the signal that carries data, *_valid is used to indicate whether the data is ready to be sent, *_flit_cancel indicates that when a flit needs to be canceled or retransmitted, this signal is used to notify other layers, and *_irdy indicates whether the data is ready to be received. Figure 4 In the example, F0Bytes[63:0], F0Bytes[127:64], etc. represent the specific bit range of the data bytes in Flit. For example, F0Bytes[63:0] represents the first 64 bits of data in Flit.

[0048] And reference Figure 5 In the protocol layer, a data buffer area can be added. The data buffer area can be a data buffer FIFO (First-in-First-out) queue. The purpose of the data buffer area is to temporarily store received data packets that the protocol layer or the upper layer of the protocol layer has no time to take away or process. As mentioned above, in the mode of transmitting user-defined data streams, the data buffer area may overflow due to the lack of a back pressure mechanism in the receiving direction, and the overflow of the data buffer area may cause the loss of received data.

[0049] In order to solve this problem, the inventor hopes to provide a data back pressure mechanism in the data receiving direction based on the adaptation layer to provide support for the data back pressure function when the user transmits a custom data stream, thereby avoiding data loss caused by data overflow in the data buffer area of ​​the receiving end when the user transmits a custom data stream.

[0050] Specifically, the inventors have found through studying the 68B flit mode and 256B flit mode supported by the UCIE protocol that the data packet format is generally a flit header (abbreviated as hdr) + data format. Taking the data packet in the 256B flit mode as an example, the format of the packet header of the data packet is as follows: Figure 6 shown. Figure 6 In the data packet header, the data packet header includes two bytes (Byte0 and Byte1), each byte includes 8 bits. The [7:6] field of Byte0 has different descriptions in the CXL3.0 and PCIe6.0 protocols. The [3:0] field of Byte0 and the [5:4] field and [3:0] field of Byte1 are related to data retransmission. The [3:0] of Byte 0 and the [3:0] of Byte 1 are combined into a sequence value, which is 8 bits in total. The [5:4] field of Byte 1 indicates: which is the corresponding sequence value: ACK (Acknowledgement, confirmation signal), NAK (NegativeAcknowledgement, negative confirmation signal) or the sequence value ES (Explicit Sequence Number) of the data itself. Data retransmission in both the data sending direction and the data receiving direction can be completed in the adaptation layer. In the related technology, the adaptation layer structure is such as Figure 7 shown.

[0051] exist Figure 7 In the figure, NEXT_TRANSMIT_SEQ Counter (next transmission sequence value counter), referred to as NTS counter, can be an eight-bit counter with a reset value of 1. The counter can perform an increment operation. When the count of the counter reaches the maximum value (for example, 255), the increment operation will become 1 (Roll Over).

[0052] REPLAY_TIMER is a retransmission timer. When this timer is raised, it indicates that the sender has sent one or more PLFs (Protocol layer flits) but has not received an acknowledgment signal (ACK / NAK) from the receiver. At this time, the sender will check the PLF copies in the retransmission buffer (PLF1 copy, PLF0 copy, etc.) and restart the retransmission timer. As long as the sender sends any PLF, the timer will start and will continue to run until the acknowledgment signal from the receiver is received. After the acknowledgment signal is received, the timer can be cleared. At this time, if there are still PLF copies in the retransmission buffer (indicating that there are still PLFs sent, but no acknowledgment signal has been received), the timer will be restarted. If the retransmission buffer is empty, the timer will not be restarted until a new PLF is sent.

[0053] REPLAY_NUM Count is a counter used to record the number of transmission failures of the same PLF. If the number of failures is too many, link retraining can be triggered.

[0054] The ACKD_SEQ register is used to store the sequence value carried by the most recently received ACK or NAK.

[0055] NEXT_RCV_SEQ is an 8-bit counter, namely, Next Receive Sequence Number, whose value is the sequence value of the successfully received PLF plus 1, and is mainly used to check whether the currently received PLF is the PLF that should be received.

[0056] NAK_SCHEDULED is a flag that is set when the receiver generates a NAK. It is cleared when the receiver successfully receives a valid PLF.

[0057] The AckNak_LATENCY_TIMER timer runs after the receiver successfully receives a valid PLF and before returning an ACK to the sender. When the AckNak_LATENCY_TIMER timer overflows, the receiver immediately returns an ACK to the sender. The sequence value carried by the ACK is NRS-1 (Next Receive Sequence minus one), that is, one ACK can indicate that all data packets with a sequence value before NRS-1 have been successfully received, achieving the purpose of one ACK corresponding to multiple valid PLFs.

[0058] For the specific process of data sending and receiving, please refer to Figure 7 , which will not be described in detail in this manual. Figure 7 It can be found that in the data sending direction, if the retransmission buffer of the adaptation layer is full, data back pressure in the data sending direction can be achieved by feeding back to the protocol layer not to send new data. However, in the data receiving direction, as long as the adaptation layer sends data to the protocol layer, the protocol layer can only receive it, and there is a risk of data buffer overflow in the protocol layer.

[0059] In order to solve this problem, the inventors proposed a waiting confirmation signal in the ACK / NAK response mechanism in the related art, which can be triggered by the receiving end protocol layer (hereinafter referred to as the first protocol layer) by pulling down the fdi_lp_trdy signal (hereinafter referred to as the first flag signal), that is, when the value of the first flag signal is the first target value (for example, it can include a low level), the adaptation layer of the receiving end (hereinafter referred to as the first adaptation layer) discards the current data packet sent by the sending end, returns the waiting confirmation signal carrying the received sequence value to the adaptation layer of the sending end (hereinafter referred to as the second adaptation layer), stops and resets the first timer (the first counter can include, for example, the AckNak_LATENCY_TIMER timer), and the waiting confirmation signal is used to instruct the second adaptation layer to control the retransmission buffer to resend the current data packet according to the received sequence value, and stops and resets the second timer (for example, it can include REPLAY_TIMER). In this way, data back pressure actively initiated by the receiving end can be realized, and by stopping the corresponding timer, the receiving end is prevented from mistakenly thinking that there is a problem with the link and retraining the link, thereby improving communication efficiency.

[0060] And, reference Figure 6 The first flag signal can utilize the reserved bit [5:4] of the data packet header Byte1 to indicate that it does not occupy the location of the data in the data packet, simplifies the data complexity, and is beneficial to improving the bandwidth utilization in the real data transmission process.

[0061] Specifically, in the data transmission method, after receiving the current data packet sent by the second core particle, the first adaptation layer determines whether to perform data back pressure according to whether the value of the first flag signal is the first target value, so as to avoid receiving a large number of data packets to the first protocol layer and causing data packet overflow or loss. Specifically, when the value of the first flag signal is the first target value, it indicates that the storage space of the data cache queue in the first protocol layer is full. At this time, the number of data packets waiting to be processed by the first protocol layer is large. If data packets are continuously received, the data packets received by the first protocol layer may overflow or be lost. Therefore, when the value of the first flag signal is the first target value, the first adaptation layer discards the current data packet received, stops and resets the first timer, and avoids the erroneous sending of the response signal; at the same time, the first adaptation layer returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer to instruct the second adaptation layer to control the retransmission buffer to resend the current data packet according to the received sequence value, so that when the first protocol layer can process the data packet normally, the first adaptation layer can receive the data packet in time and transmit it to the first protocol layer; at the same time, the waiting confirmation signal also instructs the second adaptation layer to stop and reset the second timer, so that the second adaptation layer clearly knows that there is no data packet error or link transmission error when the waiting confirmation signal is received, avoiding the reset or retraining of the data link when the second timer frequently times out the sending time. In this way, the data transmission method realizes data back pressure actively initiated by the receiving end (the first core particle) when data packets are transmitted between core particles, and at the same time avoids the incorrect resetting or retraining of the data link when the sending end (the second core particle) frequently and repeatedly sends the same data packet under the data back pressure state, thereby avoiding the reduction of data transmission efficiency caused by invalid resetting and invalid retraining of the data link in this case, which is beneficial to improving the data transmission efficiency of the data transmission method.

[0062] Based on the above concept, an embodiment of the present specification provides a data transmission method. The data transmission method provided in the embodiment of the present specification will be exemplarily described below in conjunction with the accompanying drawings.

[0063] Exemplary Methods

[0064] To be applied to Figure 8Taking the first coregrain 10 of the system on chip 100 as an example, some embodiments of this specification exemplarily illustrate the data transmission method, the first coregrain 10 includes a first protocol layer 11 and a first universal coregrain interconnect technology UCIE module, the first protocol layer 11 includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module 12 includes a first adaptation layer 121, the first adaptation layer 121 includes a first timer, the system on chip 100 also includes a second coregrain 20, the second coregrain 20 includes a second UCIE module 22, the second UCIE module 22 includes a second adaptation layer 221, the second adaptation layer 221 includes a retransmission buffer and a second timer, the data transmission method includes:

[0065] S701: In response to the current data packet sent by the second core particle 20, when the value of the first flag signal is the first target value, the first adaptation layer 121 discards the current data packet, returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer 221, stops and resets the first timer, and the waiting confirmation signal is used to instruct the second adaptation layer 221 to control the retransmission buffer to resend the current data packet according to the received sequence value, and stops and resets the second timer;

[0066] The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the first chiplet 10 has successfully received the data packet, and the first counter and the second timer are used for timing of the data retransmission mechanism.

[0067] In the implementation of the present specification, the system on chip 100 may include two or more cores, and the two cores may communicate with each other based on the UCIE technology. In the communication process between the two cores, the core that receives data may be referred to as the first core 10, and the core that sends data may be referred to as the second core 20. It can be understood that the first core 10 and the second core 20 are only used to distinguish the receiving end and the sending end in a certain data sending process. In the next data sending process, the receiving end and the sending end of the data may be interchanged. Therefore, a core in the system on chip 100 may not always receive data as the first core 10, nor may it always send data as the second core 20. It is more likely that a core in the system on chip 100 receives data as the first core 10 at a certain moment, and sends data as the second core 20 at another moment. This specification does not limit this, and it depends on the data sending and receiving requirements of the core.

[0068] In the UCIE module, the physical layer interacts directly with the hardware and is responsible for sending and receiving data streams. The physical layer can include the logical physical layer and the electrical physical layer. The logical physical layer defines the encoding method, timing and protocol of the signal, while the electrical physical layer focuses on the electrical characteristics of the signal, such as voltage level, signal integrity, impedance matching, etc., to ensure the stability and reliability of signal transmission between core particles.

[0069] The role of the adapter layer can include converting high-level data packets of the protocol layer into a format suitable for transmission at the physical layer, and decoding the data packets of the physical layer into a form that the protocol layer can understand. The adapter layer is responsible for encapsulation and decapsulation of data, and performs error detection and correction (CRC (Cyclic Redundancy Check)) when necessary.

[0070] The protocol layer defines the rules and semantics of inter-chip communication, including how to process data packets, how to respond to errors, and how to manage links. In UCIE technology, the protocol layer can include the link layer and the transaction layer.

[0071] The Link Layer is responsible for establishing, managing, and terminating the physical connection between cores. It handles the serialization and deserialization of data packets, performs error control, and provides link training and state management functions. The Link Layer ensures that data is transmitted correctly and orderly between cores, and can handle dynamic changes in the link, such as negotiation and adaptation of link speed.

[0072] The Transaction Layer is responsible for handling higher-level communication protocols, such as PCIe or CXL, which converts application data into a format that can be processed by the Link Layer. The Transaction Layer is responsible for the initiation, completion, and management of transactions, including matching requests and responses, as well as flow control and priority management.

[0073] As mentioned above, in order to distinguish the sending end from the receiving end, the sending end is referred to as the second core particle 20, the receiving end is referred to as the first core particle 10, the protocol layer, adaptation layer and physical layer that may be included in the first core particle 10 are referred to as the first protocol layer 11, the first adaptation layer 121 and the first physical layer 122 respectively, and the protocol layer, adaptation layer and physical layer that may be included in the second core particle 20 are referred to as the second protocol layer 21, the second adaptation layer 221 and the second physical layer 222 respectively, where "first" and "second" are only used to distinguish different core particles and the protocol layers, adaptation layers and physical layers in the core particles, and are not used to indicate differences in structure and function.

[0074] The first timer and the second timer can be timings for the data retransmission mechanism in the first core particle 10 and the second core particle 20. For example, the first timer can be used for the timeout timing of the response signal (ACK and NAK), and the second timer can be used for the timeout timing of the data retransmission. The first timer and the second timer cooperate to implement the data retransmission mechanism of the second core particle 20. Specifically, in an optional embodiment, the first timer is specifically used for the timeout timing of the response signal, and sends the response signal to the second adaptation layer 221 when the timing overflows. The second timer is specifically used for timing the sending duration. When the timing of the sending duration overflows, the retransmission buffer is instructed to resend the data packet cached in the retransmission buffer to the first adaptation layer 121. The sending time includes the time between the second adaptation layer 221 sending the data packet and receiving the response signal. In an optional embodiment, the first timer may include an AckNak_LATENCY_TIMER timer, and the second timer may include a REPLAY_TIMER.

[0075] In this embodiment, in addition to the response signal used for the data retransmission mechanism in the related art, a waiting confirmation signal for the receiving end to actively initiate the data back pressure function is also provided, and the waiting confirmation signal can be triggered by the first protocol layer 11 setting the first flag signal (for example, setting the value of the first flag signal to the first target value). In an optional embodiment, the first target value may include a low level. The first flag signal can be represented by the reserved bit [5:4] of the data packet header Byte1, and does not occupy the position where the data in the data packet is located.

[0076] It can be understood that the first protocol layer 11 can set the value of the first flag signal to the first target value when the data cache queue in the first protocol layer 11 is full of cached data packets or the remaining storage space in the data cache queue is less than a certain value, so as to trigger the first adaptation layer 121 to return a waiting confirmation signal carrying the received sequence value to the second adaptation layer 221, thereby realizing data back pressure actively initiated by the receiving end.

[0077] The waiting confirmation signal can instruct the second adaptation layer 221 to control the retransmission buffer to resend the current data packet according to the received sequence value, stop and reset the second timer, and on the basis of realizing the data retransmission function, stop and reset the second timer to avoid the second timer of the second core particle 20 frequently timing out, causing the second core particle 20 to mistakenly believe that there is a problem with the link and retrain the link, thereby avoiding the situation where invalid link retraining delays normal communication and improving communication efficiency.

[0078] In general, after receiving the current data packet sent by the second core particle 20, the first adaptation layer 121 determines whether to perform data back pressure according to whether the value of the first flag signal is the first target value, so as to avoid receiving a large number of data packets to the first protocol layer 11 and causing data packet overflow or loss. Specifically, when the value of the first flag signal is the first target value, it indicates that the storage space of the data cache queue in the first protocol layer 11 is full. At this time, the number of data packets waiting to be processed by the first protocol layer 11 is large. If data packets are continuously received, the data packets received by the first protocol layer 11 may overflow or be lost. Therefore, when the value of the first flag signal is the first target value, the first adaptation layer 121 discards the current data packet received, stops and resets the first timer, and avoids the erroneous sending of the response signal; at the same time, the first adaptation layer 121 returns a waiting confirmation signal carrying the received sequence value to the second adaptation layer 221. , to instruct the second adaptation layer 221 to control the retransmission buffer to resend the current data packet according to the received sequence value, so that when the first protocol layer 11 can process the data packet normally, the first adaptation layer 121 can receive the data packet in time and transmit it to the first protocol layer 11; at the same time, the waiting confirmation signal also instructs the second adaptation layer 221 to stop and reset the second timer, so that the second adaptation layer 221 knows that there is no data packet error or link transmission error when the waiting confirmation signal is received, avoiding the reset or retraining of the data link when the second timer frequently times out the transmission time. In this way, the data transmission method realizes the data back pressure actively initiated by the receiving end (the first core particle 10) during the data packet transmission between core particles, and avoids the mis-reset or mis-retraining of the data link when the sending end (the second core particle 20) frequently sends the same data packet under the data back pressure state, avoiding the reduction of data transmission efficiency caused by invalid reset and invalid retraining of the data link in this case, which is conducive to improving the data transmission efficiency of the data transmission method.

[0079] In an optional embodiment, the second core particle 20 further includes: a second protocol layer, and the second adaptation layer 221 further includes: a third timer, the third timer is used to record the waiting time, and when the waiting time times out, an error message is sent to the second protocol layer, the error message is used to indicate that the data packet receiving function of the first core particle 10 is wrong; the waiting time is used to indicate the time length from the second adaptation layer 221 receiving the waiting confirmation signal to receiving the response signal;

[0080] The waiting confirmation signal is also used to enable the third timer to count.

[0081] The third timer is used for the receiving end to record the waiting time, and the waiting time is used to represent the time from the second adaptation layer 221 receiving the waiting confirmation signal to receiving the response signal, that is, it is used to represent the time taken by the first protocol layer 11 in the first core particle 10 to release the situation in which the data cache queue is full (or occupies most of the space) with data packets. It can be understood that when the first protocol layer 11 of the first core particle 10 sends a waiting confirmation signal, it indicates that the data cache queue in the first protocol layer 11 is full (or occupies most of the space) with cached data packets, and data back pressure needs to be initiated. After that, after the second adaptation layer 221 receives the waiting confirmation signal, it will enable the third timer to start timing, and the first adaptation layer 121 will discard all the data packets received from the second core particle 20 until the first protocol layer 11 processes at least part of the cached data packets in the data cache queue. The first protocol layer 11 relieves the space shortage in the data cache queue and resets the first flag signal to a non-first target value. At this time, after receiving the current data packet sent by the second core particle 20, the first adaptation layer 121 can successfully send the current data packet to the first protocol layer 11 after the current data packet is checked and passed, and return a response signal to the second core particle 20. After receiving the response signal, the second core particle 20 can reset the third timer and stop the timing of the third timer. In this way, the time recorded by the third timer represents the duration of time that the first protocol layer 11 processes the cached data packets in the data cache queue and resolves the space shortage problem in the data cache queue. If this duration (i.e., the waiting time) is too long, it may represent that an abnormality has occurred in the first chip 10. Therefore, when the waiting time expires, an error message representing an error in the data packet receiving function of the first chip 10 may be sent to the second protocol layer to prompt the operation and maintenance personnel to handle the abnormal situation.

[0082] In an optional implementation, the response signal includes one of a confirmation signal carrying the received sequence value and a negative confirmation signal carrying the received sequence value, the confirmation signal is used to indicate that the first adaptation layer 121 successfully receives the data packet represented by the received sequence value, and the negative confirmation signal is used to indicate that the first adaptation layer 121 fails to successfully receive the current data packet;

[0083] The data transmission method further comprises:

[0084] The first adaptation layer 121 verifies the current data packet in response to the current data packet sent by the second core particle 20, and sets the second flag signal to the second target value when the verification fails;

[0085] When the check passes, the relationship between the sequence value of the current data packet and the target sequence value is determined, the received sequence value is equal to the target sequence value minus one, and when the sequence value of the current data packet is equal to the target sequence value and the value of the first flag signal is not the first target value, the current data packet is sent to the first protocol layer 11 to successfully receive the current data packet;

[0086] When the sequence value of the current data packet is equal to the target sequence value and the value of the first flag signal is the first target value, discard the current data packet and set the value of the first flag signal to the first target value;

[0087] When the sequence value of the current data packet is greater than the target sequence value, the second flag signal is set to the second target value; the second flag signal having the second target value is used to indicate that the first adaptation layer 121 has received an erroneous current data packet;

[0088] When the sequence value of the current data packet is less than the target sequence value, generating the confirmation signal;

[0089] Based on the first flag signal, the second flag signal and the confirmation signal, the response signal or the waiting confirmation signal is generated and sent to the second adaptation layer 221 .

[0090] Specifically, in an optional implementation, the generating the response signal or the waiting confirmation signal based on the first flag signal, the second flag signal and the confirmation signal and sending it to the second adaptation layer 221 includes:

[0091] When the value of the first flag signal is the first target value, generating the waiting confirmation signal and sending it to the second adaptation layer 221;

[0092] When the value of the first flag signal is not the first target value and the value of the second flag signal is the second target value, generating the negative confirmation signal and sending it to the second adaptation layer 221;

[0093] When the value of the first flag signal is not the first target value and the confirmation signal exists, the confirmation signal is sent to the second adaptation layer 221 .

[0094] In this embodiment, a mechanism for generating a response signal and a waiting confirmation signal is provided. Based on this mechanism, the generation requirement of the waiting confirmation signal can be met while meeting the generation requirement of the response signal when data back pressure needs to be initiated at the receiving end. The logic adjustment of the first adaptation layer 121 is small, and the overall logic is clear. The generation of the response signal and the waiting confirmation signal does not interfere with each other, which is conducive to ensuring the normal operation of the method.

[0095] In order to ensure the normal operation of the data back pressure function initiated by the receiving end, and to avoid the second adaptation layer 221 not stopping and resetting the first timer due to the loss of the waiting confirmation signal, thereby causing the second adaptation layer 221 to mistakenly retrain the link, in one embodiment, the data transmission method further includes:

[0096] resending the waiting confirmation signal when the interval between the last sending of the waiting confirmation signal exceeds a predetermined interval;

[0097] When the value of the first flag signal changes from the first target value to a non-first target value, the value of the first flag signal is set to a non-first target value.

[0098] In this embodiment, when the interval from the last sending of the waiting confirmation signal exceeds the predetermined interval, in order to ensure that the second chip 20 can successfully receive the waiting confirmation signal, the waiting confirmation signal can be resent when the value of the first flag signal reaches the first target value to ensure that the second chip 20 can normally receive the waiting confirmation signal.

[0099] In a specific implementation, a feasible data transmission method execution process is provided, referring to Fig. 9 , Fig. 9 The left side shows the protocol layer and adaptation layer in the sending end, and the right side shows the protocol layer and adaptation layer in the receiving end. Fig. 9 It mainly shows the processing performed by the adaptation layer at the sending end and the adaptation layer at the receiving end during the data packet sending process.

[0100] When the sender sends data, the protocol layer of the sender sends the data packet PLFs to the adaptation layer of the sender. The adaptation layer adds a sequence value to the protocol layer data packet, and the NEXT_TRANSMIT_SEQ (NTS) counter performs an increment operation to record the sequence value that should be added to the next data packet. When adding the sequence value, if an acknowledgment signal (ACK / NAK) or a waiting confirmation signal (WACK) is received from the receiving end, the sequence value type is marked with the acknowledgment signal or the waiting confirmation signal (the acknowledgment signal or the waiting confirmation signal is written into the hdr of the data packet). If the acknowledgment signal (ACK / NAK) or the waiting confirmation signal (WACK) is not received from the receiving end, the sequence value type is marked as ES.

[0101] Generate CRC for verification at the receiving end, and put the data packet to be sent into the retransmission buffer. After sending the data packet to the receiving end through the physical layer of the sending end, the retransmission buffer will save the data packet copy (PLF0 copy, PLF1 copy, etc.) for the data retransmission mechanism. If the retransmission buffer is full, the protocol layer of the sending end is fed back to not send new data, thereby realizing data back pressure at the sending end.

[0102] After the adaptation layer of the receiving end receives the current data packet (hereinafter referred to as data packet) transmitted by the physical layer of the receiving end, the CRC check can be performed on the data packet. If the check fails, the second flag signal is set to the second target value, indicating that the receiving end has received an erroneous data packet, and the ACK / NAK / WACK generator generates a NAK and returns it to the receiving end;

[0103] If the check passes, the sequence value type (ES / ACK / NAK / WACK) is determined by the hdr of the data packet, where ES, ACK and NAK have been introduced in the previous article and will not be repeated here. WACK refers to waiting for the confirmation signal; if it is the ES type, the relationship between the received sequence value and NRS is determined, and NRS is the target sequence value mentioned above. If the sequence value of the data packet is less than the target sequence value, ACK is generated and returned to the adaptation layer of the sender. If the sequence value of the data packet is equal to the target sequence value, it indicates that the data packet is a data packet that needs to be received. At this time, the NAK flag (NAK_SCHEDULED, that is, the second flag signal is set to a non-second target value) can be cleared, and combined with the sequence value type judgment result, it is determined whether fdi_lp_trdy is valid (that is, whether the first flag signal is a non-first target value). If so, the data is correct and can be sent to the protocol layer of the receiving end, and the count value of the NEXT_RCV_SEQ (NRS) counter is increased by 1 (NRS+1) to update the target sequence value NRS;

[0104] If not, the data packet is discarded and the WACK flag is set (WACK_SCHEDULED, that is, the first flag signal is set to the first target value). The ACK / NAK / WACK generator generates a response signal (ACK / NAK) or a waiting confirmation signal (WACK) based on the first flag signal, the second flag signal and the confirmation signal (ACK) and sends it to the adaptation layer of the sending end.

[0105] If the sequence value in hdr is marked as ACK / NAK / WACK type, the data packet can be called a management data packet, and the data packet is returned to the adaptation layer of the sender. The adaptation layer of the sender determines whether the sequence value of the received ACK / NAK / WACK is equal to AS (that is, the sequence value of the most recently received ACK or NAK stored in the ACKD_SEQ register). If not, the ACKD_SEQ register is updated, and the old PLF data packet in the retransmission buffer is cleared according to the received sequence value (because the received sequence value represents the maximum value of the sequence value of the data packet already received by the receiving end, the data packets before the sequence value have been successfully received and do not need to be resent, so the PLF copy before the sequence value in the retransmission buffer can be cleared). After updating the ACKD_SEQ register, RETPLAY_NUM and REPLAY_TIMER (the first timer) are cleared to determine whether (NTS-AS) is greater than or equal to 128. If so, an error is reported and data transmission is stopped. If not, the data packet transmission process continues.

[0106] When the received sequence value type is NAK / WACK, or when the sequence value carried by the received data packet = AS, determine whether the sequence value type is NAK or WACK? If so, retransmit the data packet carrying the received sequence value carried by NAK or WACK, and when the sequence value type is WACK, stop and reset RETPLAY_NUM and REPLAY_TIMER (the first timer), and when the sequence value type is not WACK, RETPLAY_NUM counter +1 (indicating the number of failed data packet transmissions +1).

[0107] In addition, a third timer (WACK_TIMER) is set in the adaptation layer of the sending end. When the sequence value type of the received data packet is WACK, the third timer is enabled. When the third timer reaches the threshold time, an error is reported to the protocol layer of the sending end (that is, when the waiting time recorded by the third timer times out, an error message is sent to the second protocol layer).

[0108] In the adaptation layer of the receiving end, a first timer (AckNak_LATENCY_TIMER timer, or ACK / NAK delay timer) is set, which will run before the receiving end successfully receives a valid PLF and does not return an ACK to the sending end. When the AckNak_LATENCY_TIMER timer overflows, the receiving end can return an ACK to the sending end, and the sequence value carried by the ACK is NRS-1 (Next Receive Sequence minus one), that is, one ACK can indicate that all data packets with a sequence value before NRS-1 are successfully received, achieving the purpose of one ACK corresponding to multiple valid PLFs.

[0109] Below Fig.10 , Fig.11 and Fig.12 , respectively introduces the ACK, NAK and WACK mechanisms. Figure 10 to Figure 12 In the figure, dieA represents the data sending end, and dieB represents the data receiving end. Figure 10 to Figure 12 The physical layer is not shown in the figure. The protocol layer in dieA is the second protocol layer mentioned above, the adaptation layer in dieA is the second adaptation layer 221 mentioned above, the protocol layer in dieB is the first protocol layer 11 mentioned above, and the adaptation layer in dieB is the first adaptation layer 121 mentioned above.

[0110] exist Fig.10 The ACK mechanism is shown in Figure 1, which is the process when the receiving end successfully receives the data packet:

[0111] Step 1 dieA is ready to send 5 PLFs to dieB in sequence, and the corresponding sequence values ​​are 3, 4, 5, 6, and 7 respectively; assuming that the fdi_lp_trdy signal is always in a valid state;

[0112] Step 2 dieB successfully receives PLF3 and increases NEXT_RCV_SEQ from 3 to 4, but dieB does not immediately return Ack to dieA (AckNak_LATENCY_TIMER has not overflowed yet);

[0113] Step3 dieB successfully receives PLF4 and PLF5;

[0114] Step 4: Assuming that AckNak_LATENCY_TIMER overflows at this time, dieB returns an ACK with a sequence value of 5 to dieA. At the same time, dieB resets AckNak_LATENCY_TIMER, but does not restart it until dieB successfully receives PLF6.

[0115] Step 5 dieA receives Ack5, resets REPLAY_TIMER and REPLAY_NUM, and then removes the PLF backup with sequence value 5 (and before 5) in the retransmission buffer;

[0116] Step 6 Once dieB receives PLF6, AckNak_LATENCY_TIMER will be restarted.

[0117] exist Fig.11 The NAK mechanism is shown in Figure 1, which is the process of processing when the receiving end fails to successfully receive the data packet:

[0118] Step 1: Assume that dieA is going to send PLFs with sequence values ​​of 254, 255, 1, 2, and 3 to dieB in sequence; Assume that the fdi_lp_trdy signal is always in a valid state;

[0119] Step 2 dieB successfully accepts PLF 254 and increases NEXT_RCV_SEQ by 1 to 255;

[0120] Step 3 dieB receives PLF 255, but the PLF does not pass the CRC check (i.e., there is an error). At this time, no matter what state AckNak_LATENCY_TIMER is in, dieB will return Nak 254 to dieA (note that the sequence value of the returned Nak is the sequence value of the last successfully received PLF, i.e., the sequence value of NAK is NRS-1). At the same time, dieB stops and resets AckNak_LATENCY_TIMER;

[0121] Step 4 dieB will keep waiting for dieA to send PLF 255 to it, but dieA does not know what happened. Before receiving the Ack / Nak returned by dieB, it will continue to send PLF1~PLF3, but dieB will directly ignore these PLFs.

[0122] Step 5 When dieA receives Nak 254 from dieB, it removes PLF 254 (and the previous PLF) in the buffer and starts sending again from PLF 255. At the same time, REPLAY_TIMER and REPLAY_NUM are reset.

[0123] Step 6: Since dieA receives Nak instead of Ack, dieA restarts REPLAY_TIMER and increases REPLAY_NUM by one. When REPLAY_NUM overflows, it indicates that the same data packet has failed to be sent many times, and dieA can retrain the link.

[0124] Step 7: Once dieB successfully receives PLF 255, dieB clears the NAK_SCHEDULED flag, increments the NEXT_RCV_SEQ counter by one, and restarts the AckNak_LATENCY_TIMER.

[0125] exist Fig.12 The WACK mechanism is shown in Figure 1, which is the processing process when the receiving end initiates active data back pressure:

[0126] Step 1 assumes that dieA is going to send PLFs with sequence values ​​of 254, 255, 1, 2, 3 to dieB in sequence;

[0127] Step 2 dieB successfully accepts PLF 254 and increases NEXT_RCV_SEQ by 1 to 255; but afterwards, the protocol layer pulls the fdi_lp_trdy signal low, making fdi_lp_trdy invalid.

[0128] Step 3 dieB receives PLF 255. Since fdi_lp_trdy is invalid at this moment, it discards the data. At this time, no matter what state AckNak_LATENCY_TIMER is in, dieB will return WACK 254 to dieA (note that the sequence value of the returned WACK is the sequence value of the last successfully received PLF). At the same time, dieB stops and resets AckNak_LATENCY_TIMER to avoid returning ACK or NAK to dieA by mistake when the timer overflows;

[0129] Step 4 dieB will keep waiting for dieA to send PLF 255 to it, but dieA does not know what happened. Before receiving the Ack / Nak / WACK returned by dieB, it will continue to send PLF1~PLF3, but dieB will directly ignore these PLFs.

[0130] Step 5: When dieA receives WACK 254 from dieB, it removes PLF 254 (and the previous PLF) in the buffer and starts sending again from PLF 255. At the same time, REPLAY_TIMER and REPLAY_NUM are reset.

[0131] Step 6: Since dieA receives WACK, dieA will not start REPLAY_TIMER and REPLAY_NUM will not increase by 1; instead, the WACK_TIMER timer is started, thus avoiding link retraining under data back pressure due to REPLAY_NUM count overflow.

[0132] Step 7 Once dieB successfully receives PLF 255 and fdi_lp_trdy becomes valid, the protocol layer successfully obtains the data. dieB will clear the WACK_SCHEDULED flag, increase the NEXT_RCV_SEQ counter by one, and restart AckNak_LATENCY_TIMER. If fdi_lp_trdy in dieB is always invalid, dieB can send WACK repeatedly. During this period, even if NAK_SCHEDULED is 1, only WACK is sent. When the WACK_TIMER timer in dieA reaches the threshold, an error is reported to the protocol layer of dieA. Due to the back pressure of dieB, the data of dieA cannot be sent for a long time. The threshold of the WACK_TIMER timer can be set according to the overall link delay and the system's requirements for data stability, which can generally be at the ms (millisecond) level.

[0133] Accordingly, reference Fig.13 Taking the second coregrain 20 applied to the system on chip 100 as an example, the embodiment of this specification also provides a data transmission method, the second coregrain 20 includes a second UCIE module 22, the second UCIE module 22 includes a second adaptation layer 221, the second adaptation layer 221 includes a retransmission buffer and a second timer, the system on chip 100 also includes: a first coregrain 10, the first coregrain 10 includes a first protocol layer 11 and a first UCIE module 12, the first protocol layer 11 includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module 12 includes a first adaptation layer 121, the first adaptation layer 121 includes a first timer; the data transmission method includes:

[0134] S1301: The second adaptation layer 221 sends a current data packet to the first adaptation layer 121, where the current data packet is used to indicate that the first adaptation layer 121, in response to the current data packet sent by the second core particle 20, discards the current data packet when the value of the first flag signal is the first target value, returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer 221, and stops and resets the first timer;

[0135] S1302: The second adaptation layer 221 controls the retransmission buffer to resend the current data packet according to the received sequence value in response to the waiting confirmation signal, and stops and resets the second timer;

[0136] The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the first chiplet 10 has successfully received the data packet, and the first counter and the second timer are used for timing of the data retransmission mechanism.

[0137] Optionally, the first timer is used for timing the timeout of the response signal, and sends the response signal to the second adaptation layer 221 when the timing overflows. The second timer is used for timing the sending duration. When the timing of the sending duration overflows, it instructs the retransmission buffer to resend the data packets cached in the retransmission buffer to the first adaptation layer 121. The sending time includes the time between the second adaptation layer 221 sending the data packet and receiving the response signal.

[0138] Optionally, the data transmission method further includes:

[0139] In response to the waiting confirmation signal, the second adaptation layer 221 deletes the data packets whose sequence values ​​cached in the retransmission buffer are located before the received sequence value.

[0140] Optionally, the second adaptation layer 221 further includes: a timeout counter;

[0141] The data transmission method further comprises:

[0142] The timeout counter records the number of retransmissions of the same data packet, and when the number of retransmissions exceeds a preset threshold, the target link is reset or retrained, and the target link includes the communication link between the first adaptation layer 121 and the second adaptation layer 221 .

[0143] The timeout counter may include, for example, the RETPLAY_NUM mentioned above.

[0144] Optionally, the data transmission method further includes:

[0145] The second adaptation layer 221 writes the waiting confirmation signal into the header of the management data packet, and sends the management data packet to the first adaptation layer 121, wherein the management data packet is used to instruct the first adaptation layer 121 to return the management data packet to the second adaptation layer 221;

[0146] The second adaptation layer 221 parses the waiting confirmation signal in the management data packet, and executes the steps of controlling the retransmission buffer to resend the current data packet according to the received sequence value in response to the waiting confirmation signal, and stopping and resetting the second timer.

[0147] In this implementation manner, the management data packet may refer to a data packet whose sequence value type identified in the packet header is a response signal or a waiting confirmation signal.

[0148] For the specific limitations of the second chip 20 executing the data transmission method, please refer to the above description, which will not be elaborated in this specification.

[0149] Exemplary Devices

[0150] like Fig.14 As shown, in an optional embodiment, a system on chip 100 is also provided, including: a first core particle 10 and a second core particle 20; the first core particle 10 includes a first protocol layer 11 and a first UCIE module 12, the first protocol layer 11 includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module 12 includes a first adaptation layer 121, the first adaptation layer 121 includes a first timer, the second core particle 20 includes a second UCIE module 22, the second UCIE module 22 includes a second adaptation layer 221, the second adaptation layer 221 includes a retransmission buffer and a second timer; wherein,

[0151] The second adaptation layer 221 is configured to: send the current data packet to the first adaptation layer 121, in response to the waiting confirmation signal, control the retransmission buffer to resend the current data packet according to the received sequence value carried by the waiting confirmation signal, and stop and reset the second timer;

[0152] The first adaptation layer 121 is configured to: in response to the current data packet sent by the second core particle 20, when the value of the first flag signal is the first target value, discard the current data packet, return a waiting confirmation signal carrying the received sequence value to the second adaptation layer, and stop and reset the first timer;

[0153] The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

[0154] exist Fig.14, the first physical layer 122 in the first UCIE module 12, the second physical layer 222 in the second UCIE module 22, and the second protocol layer 21 in the second core particle 20 are also shown. For the relevant definitions of each hardware module in the system on chip 100, reference may be made to the relevant description above.

[0155] Another embodiment of this specification also provides a computing device, see Fig.15 As shown, an exemplary embodiment of the present specification also provides a computing device, including: a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the data transmission method according to various embodiments of the present specification described in the above embodiments of the present specification are executed.

[0156] The internal structure of the computing device can be as follows Fig.15 As shown, the computing device includes a processor, a memory, a network interface and an input device connected through a system bus. Among them, the processor of the computing device is used to provide computing and control capabilities. The memory of the computing device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computing device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps of the data transmission method according to various embodiments of the present specification described in the above embodiments of the present specification are performed.

[0157] The processor may include a main processor and may also include a baseband chip, a modem, etc.

[0158] The memory stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes computer operation instructions. More specifically, the memory may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.

[0159] The processor may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the scheme of the present invention. It may also be a digital signal processor (DSP), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component.

[0160] The input device may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0161] Output devices may include means that allow information to be output to a user, such as display screens, printers, speakers, etc.

[0162] The communication interface may include using any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0163] The processor executes the program stored in the memory and calls other devices, which can be used to implement each step of any data transmission method provided in the above embodiments of this specification.

[0164] The computing device may also include a display component and a voice component. The display component may be a liquid crystal display or an electronic ink display. The input device of the computing device may be a touch layer covered on the display component, or a button, trackball or touchpad provided on the housing of the computing device, or an external keyboard, touchpad or mouse.

[0165] Those skilled in the art will understand that Fig.15 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of this specification, and does not constitute a limitation on the computing device to which the scheme of this specification is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0166] Exemplary computer program products and storage media

[0167] In addition to the above-mentioned methods and devices, the data transmission method provided in the embodiments of this specification may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the data transmission method according to various embodiments of this specification described in the above "Exemplary Method" section of this specification.

[0168] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is implemented as a computer storage medium. In another optional embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0169] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the embodiments of the present specification, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0170] In addition, an embodiment of the present specification also provides a computer-readable storage medium on which a computer program is stored, and the computer program is executed by a processor to execute the steps of the data transmission method according to various embodiments of the present specification described in the above "Exemplary Method" section of the present specification.

[0171] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this specification can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0172] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0173] The above-mentioned embodiments only express several implementation methods of this specification, and the descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the solutions provided by the embodiments of this specification. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of this specification, which all belong to the protection scope of this specification. Therefore, the protection scope of the patent of this specification shall be based on the attached claims.

Claims

1. A data transmission method, characterized in that: A first coregrain applied to a system on chip, the first coregrain includes a first protocol layer and a first universal coregrain interconnect technology UCIE module, the first protocol layer includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module includes a first adaptation layer, the first adaptation layer includes a first timer, the system on chip also includes a second coregrain, the second coregrain includes a second UCIE module, the second UCIE module includes a second adaptation layer, the second adaptation layer includes a retransmission buffer and a second timer, the data transmission method includes: In response to the current data packet sent by the second core particle, when the value of the first flag signal is the first target value, the first adaptation layer discards the current data packet, returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer, stops and resets the first timer, and the waiting confirmation signal is used to instruct the second adaptation layer to control the retransmission buffer to resend the current data packet according to the received sequence value, and stops and resets the second timer; The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

2. The method according to claim 1, characterized in that The first timer is specifically used for timing the timeout of the response signal. When the timing overflows, the response signal is sent to the second adaptation layer. The second timer is specifically used for timing the sending duration. When the timing of the sending duration overflows, the retransmission buffer is instructed to resend the data packet cached in the retransmission buffer to the first adaptation layer. The sending time includes the time from the second adaptation layer sending the data packet to receiving the response signal.

3. The method according to claim 2, characterized in that The second core particle further includes: a second protocol layer, and the second adaptation layer further includes: a third timer, the third timer is used to record the waiting time, and when the waiting time times out, an error message is sent to the second protocol layer, the error message is used to indicate that the data packet receiving function of the first core particle is wrong; the waiting time is used to indicate the time length from the second adaptation layer receiving the waiting confirmation signal to the second adaptation layer receiving the response signal; The waiting confirmation signal is also used to enable the third timer to count.

4. The method according to claim 3, characterized in that The response signal includes one of a confirmation signal carrying the received sequence value and a negative confirmation signal carrying the received sequence value, the confirmation signal is used to indicate that the first adaptation layer successfully receives the data packet represented by the received sequence value, and the negative confirmation signal is used to indicate that the first adaptation layer does not successfully receive the current data packet; The data transmission method further comprises: The first adaptation layer verifies the current data packet in response to the current data packet sent by the second core particle, and sets the second flag signal to a second target value when the verification fails; When the check passes, determining the relationship between the sequence value of the current data packet and the target sequence value, the received sequence value is equal to the target sequence value minus one, and when the sequence value of the current data packet is equal to the target sequence value and the value of the first flag signal is not the first target value, sending the current data packet to the first protocol layer to successfully receive the current data packet; When the sequence value of the current data packet is equal to the target sequence value and the value of the first flag signal is the first target value, discard the current data packet and set the value of the first flag signal to the first target value; When the sequence value of the current data packet is greater than the target sequence value, setting the second flag signal to the second target value; the second flag signal having the second target value is used to indicate that the first adaptation layer has received an erroneous current data packet; When the sequence value of the current data packet is less than the target sequence value, generating the confirmation signal; Based on the first flag signal, the second flag signal and the confirmation signal, the response signal or the waiting confirmation signal is generated and sent to the second adaptation layer.

5. The method according to claim 4, characterized in that The generating the response signal or the waiting confirmation signal based on the first flag signal, the second flag signal and the confirmation signal and sending it to the second adaptation layer comprises: When the value of the first flag signal is the first target value, generating the waiting confirmation signal and sending it to the second adaptation layer; When the value of the first flag signal is not the first target value and the value of the second flag signal is the second target value, generating the negative confirmation signal and sending it to the second adaptation layer; When the value of the first flag signal is not the first target value and the confirmation signal exists, the confirmation signal is sent to the second adaptation layer.

6. The method according to claim 1, characterized in that Also includes: resending the waiting confirmation signal when the interval between the last sending of the waiting confirmation signal exceeds a predetermined interval; When the value of the first flag signal changes from the first target value to a non-first target value, the value of the first flag signal is set to a non-first target value.

7. A data transmission method, characterized in that: A second coregrain applied to a system on chip, the second coregrain includes a second UCIE module, the second UCIE module includes a second adaptation layer, the second adaptation layer includes a retransmission buffer and a second timer, the system on chip also includes: a first coregrain, the first coregrain includes a first protocol layer and a first UCIE module, the first protocol layer includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module includes a first adaptation layer, the first adaptation layer includes a first timer; the data transmission method includes: The second adaptation layer sends a current data packet to the first adaptation layer, where the current data packet is used to indicate that the first adaptation layer, in response to the current data packet sent by the second core particle, discards the current data packet when the value of the first flag signal is the first target value, returns a waiting confirmation signal carrying a received sequence value to the second adaptation layer, and stops and resets the first timer; The second adaptation layer controls the retransmission buffer to resend the current data packet according to the received sequence value in response to the waiting confirmation signal, and stops and resets the second timer; The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

8. The method according to claim 7, characterized in that The first timer is used for timing the timeout of the response signal, and sends the response signal to the second adaptation layer when the timing overflows. The second timer is used for timing the sending duration. When the timing of the sending duration overflows, it instructs the retransmission buffer to resend the data packet cached in the retransmission buffer to the first adaptation layer. The sending time includes the time from the second adaptation layer sending the data packet to the receipt of the response signal.

9. The method according to claim 7, characterized in that: Also includes: In response to the waiting confirmation signal, the second adaptation layer deletes the data packet whose sequence value cached in the retransmission buffer is located before the received sequence value.

10. The method according to claim 7, characterized in that The second adaptation layer further includes: a timeout counter; The data transmission method further comprises: The number of retransmissions of the same data packet is recorded based on the timeout counter. When the number of retransmissions exceeds a preset threshold, a target link is reset or retrained. The target link includes a communication link between the first adaptation layer and the second adaptation layer.

11. The method according to claim 7, characterized in that Also includes: The second adaptation layer writes the waiting confirmation signal into the header of a management data packet, and sends the management data packet to the first adaptation layer, where the management data packet is used to instruct the first adaptation layer to return the management data packet to the second adaptation layer; The second adaptation layer parses the waiting confirmation signal in the management data packet, and executes the steps of the second adaptation layer responding to the waiting confirmation signal, controlling the retransmission buffer to resend the current data packet according to the received sequence value, and stopping and resetting the second timer.

12. A system on chip, characterized in that: include: a first core-grain and a second core-grain; the first core-grain includes a first protocol layer and a first UCIE module, the first protocol layer includes a data cache queue, the data cache queue is used to cache data packets, the first UCIE module includes a first adaptation layer, the first adaptation layer includes a first timer, the second core-grain includes a second UCIE module, the second UCIE module includes a second adaptation layer, the second adaptation layer includes a retransmission buffer and a second timer; wherein, The second adaptation layer is configured to: send the current data packet to the first adaptation layer, in response to the waiting confirmation signal, control the retransmission buffer to resend the current data packet according to the received sequence value carried by the waiting confirmation signal, and stop and reset the second timer; The first adaptation layer is configured to: in response to a current data packet sent by the second core particle, when the value of the first flag signal is the first target value, discard the current data packet, return a waiting confirmation signal carrying a received sequence value to the second adaptation layer, and stop and reset the first timer; The first flag signal having the first target value is used to indicate that the storage space of the data cache queue is full, the received sequence value is used to indicate that the data packet is successfully received by the first core particle, and the first counter and the second timer are used for timing of the data retransmission mechanism.

13. A computing device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the data transmission method according to any one of claims 1 to 11 when executing the computer program.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the data transmission method according to any one of claims 1 to 11 is implemented.

Citation Information

Cited By

  • Data transmission system and method, electronic equipment and storage medium

    CN119892722A

  • Communication method, system and chip

    CN120750498A

  • Communication methods, systems and chips

    CN120750498B