A reliable method to improve the robustness of dual-processor cross-upgrades

By using FPGA dual-port RAM and a multi-layer verification mechanism, reliable remote upgrades of the dual-DSP system were achieved, solving the problems of high hardware cost and low reliability in existing technologies, and realizing efficient upgrades without human intervention.

CN123114516APending Publication Date: 2026-11-17HUNAN GUOKE HAIPHONG INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611614547.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In dual-DSP embedded systems, existing technologies require equipping the DSP with network interfaces and network protocol stacks, which increases hardware costs, makes upgrade channels inefficient and unreliable, and makes it impossible to achieve reliable remote upgrades of unattended equipment.

Method used

By using the dual-port RAM inside the FPGA as an intermediary, a breakpoint resume and multi-layer verification mechanism are implemented through the packet number register, eliminating the need for network hardware and protocol stack from the DSP. Magic number triggering is used to enable remote upgrade triggering from the DSP without network access. Combined with the upgrade handshake register group and readback verification mechanism, the reliability of data transmission is ensured.

Benefits of technology

It significantly reduces hardware costs, enables reliable remote upgrades without on-site intervention, eliminates the need to retransmit data after communication interruption, detects and retransmits erroneous data through a multi-layer verification mechanism, prevents startup anomalies caused by program area residue, and improves the reliability and maintainability of upgrades.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114516A_ABST
    Figure CN123114516A_ABST
Patent Text Reader

Abstract

This invention discloses a reliable method for improving the robustness of cross-upgrades of dual processors, relating to the field of remote upgrade technology for embedded systems. This method uses FPGA dual-port RAM to transfer upgrade data from the master DSP to a network-less slave DSP. It utilizes a magic number trigger register and power reset to remotely enter upgrade mode from the slave DSP. A packet number register enables breakpoint resumption and retransmission alignment. A three-layer mechanism—packet number alignment, dual-port RAM readback verification, and NOR Flash word-by-word readback verification—ensures upgrade reliability. This invention eliminates the need for slave DSP network hardware and protocol stacks, supports breakpoint resumption, and is recoverable in upgrade failures, making it suitable for remote maintenance of unattended equipment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of embedded system program upgrade technology, specifically to a method for cross-remote upgrade of master and slave dual processors using FPGA dual-port RAM, which is particularly suitable for scenarios where the slave DSP has no network interface and runs on bare metal. Background Technology

[0002] In dual-DSP embedded systems, remote program upgrades are a core function of equipment maintenance. Traditional solutions require each DSP to be equipped with its own network interface (Ethernet PHY / MAC or UART adapter). The slave DSP requires additional communication hardware configuration and TCP / IP protocol stack porting, resulting in doubled hardware costs, increased PCB area, and increased power consumption. Implementing a complete network protocol stack on the slave DSP under bare-metal operation is extremely difficult, and upgrade channels often rely on low-speed serial ports or dedicated upgrade chips. The unified burning method after full file transfer requires retransmission from the beginning after communication interruption. Large-capacity firmware upgrades are time-consuming and lack dual protection of transmission verification and Flash readback verification. Communication errors or NOR Flash write failures cannot be detected in time, easily leading to equipment failure. Entering upgrade mode from the slave DSP usually requires on-site operation of DIP switches or short-contact triggers, which is not conducive to unattended equipment maintenance. Summary of the Invention

[0003] This invention aims to provide a cross-upgrade method that simultaneously upgrades both master and slave DSPs, requiring only the master DSP to be connected to the network. This eliminates the need for network hardware and protocol stacks in the slave DSP, uses the internal dual-port RAM of the FPGA as a data transfer channel for upgrades, and does not add any additional communication pins. It enables breakpoint resumption through a packet number register, ensures upgrade reliability through a multi-layer verification mechanism, and enables remote upgrade triggering of the slave DSP without a network by using magic number triggering.

[0004] To address the aforementioned technical problems, this invention provides a reliable method for improving the robustness of dual-processor cross-upgrades, comprising the following steps: S100: The host computer sends upgrade data to the slave DSP, which has no network interface, via the main DSP. The main DSP and the slave DSP exchange upgrade data through the dual-port RAM inside the FPGA and perform a packet-by-packet handshake through the upgrade handshake register group. In this step, the host computer communicates with the main DSP via Ethernet TCP. The main DSP is connected to the FPGA via the EMIF asynchronous bus and exclusively occupies the remote communication resources. The slave DSP exchanges upgrade data with the main DSP through the dual-port RAM inside the same FPGA. The FPGA internally includes a dual-port RAM data buffer, a readback verification RAM, and an upgrade handshake register. The upgrade handshake register includes an upgrade trigger register, a write completion flag register, a read completion flag register, a current packet number register, a total packet count register, and an error flag register.

[0005] S200, the host computer issues an upgrade command, the main DSP writes a predetermined magic number to the upgrade trigger register and resets the power supply of the slave DSP. After the slave DSP is powered on and recognizes the predetermined magic number, it enters the upgrade mode. In this step, the predetermined magic number is 0xB00A. After the slave DSP recognizes the predetermined magic number, it enters the upgrade mode. Otherwise, it jumps to the application area through the watchdog timer.

[0006] S300 and the main DSP write the upgrade data into the dual-port RAM in fixed-length data blocks, and at the same time write the current packet number to the packet number register and set the write completion flag. In this step, the fixed-length data block is 64 16-bit words, the packet number is incremented from 0, and each packet is written to the packet number register first and then the write completion flag is set.

[0007] S400: Align the local packet number with the packet number register value from the DSP, then read the data from the dual-port RAM and copy it to the readback verification RAM, and then set the read completion flag.

[0008] The S500 and main DSP compare the readback verification RAM with the local cache. If they are inconsistent, the write completion flag is set to an error value and a retransmission is triggered. If they are consistent, the write completion flag is set and the verification is deemed successful. In this step, the slave DSP copies the data to the readback verification RAM as is while reading the data. The main DSP reads the copy from the independent address space for comparison, without occupying the slave DSP's computing resources.

[0009] S600: After the data from the DSP passes the verification, it is written to the NOR Flash and compared word by word. If it fails, the error flag is written and the process is stopped. If it matches, the completion flag is set and the main DSP is notified. In this step, the NOR Flash writing address is the program area base address plus the packet number multiplied by the packet length. If the word-by-word comparison fails, the predetermined error code 0x00EE is written.

[0010] S700: The main DSP reads the error flag and reports the upgrade result to the host computer. After all is completed, the main DSP clears the upgrade trigger register, the slave DSP is reset and jumps to the application area. In this step, the main DSP reads the error flag and reports it to the host computer. After success, it clears the trigger register to make the slave DSP jump. The final interpretation of the upgrade result is completed by the main DSP and reported to the host computer, forming a closed-loop management of the host computer, the main DSP and the slave DSP.

[0011] This invention utilizes the physical dual-port characteristics of FPGA dual-port RAM to enable the master DSP and slave DSP to complete data transmission verification in a closed-loop manner of writing, reading back and comparing. Reliable upgrades can be achieved without the slave DSP having network capabilities, thereby significantly reducing hardware costs. At the same time, the multi-layer verification mechanism enables communication errors and write failures to be detected and retransmitted.

[0012] Furthermore, the process also includes step S200': after entering upgrade mode from the DSP, reading the total number of packages register to obtain the total number of upgrade files, and pre-erasing all sectors of the program area in the NOR Flash. The NOR Flash must be erased before writing. Pre-erasing all sectors avoids the address alignment complexity and time overhead caused by erasing packages one by one, thereby ensuring that the storage area is clean before writing and preventing program area residue from causing startup abnormalities.

[0013] Furthermore, the readback verification RAM and the dual-port RAM data buffer are independently mapped regions in the FPGA address space. This independent address space allows the main DSP to read a copy of the data written by the DSP from the "other side" of the dual-port RAM, thereby avoiding access conflicts, achieving physical isolation for transmission verification, and improving reliability.

[0014] Compared with existing technologies, this invention eliminates the need for DSP network hardware and protocol stacks through an FPGA dual-port RAM intermediary architecture, significantly reducing hardware costs. It enables remote upgrade triggering from the DSP without a network by using magic number triggering and power reset, requiring no on-site intervention. Packet number register alignment enables breakpoint resumption and retransmission alignment, eliminating the need to re-transmit after communication interruption recovery. A three-layer mechanism—packet number alignment, dual-port RAM readback verification, and NOR Flash word-by-word readback verification—works collaboratively to detect, retransmit, or safely terminate communication errors and write failures. Automatic reset and retry in case of DSP upgrade anomalies, combined with pre-erasure and word-by-word readback, prevents startup anomalies caused by program area remnants. The overall solution has significant advantages in hardware cost, software simplicity, upgrade reliability, and maintainability. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the embodiments described below. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart of a reliable method for improving the robustness of dual-processor cross-upgrades according to the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to illustrate the present invention and should not be regarded as limiting the scope of the present invention.

[0017] like Figure 1 As shown, this invention provides a reliable method for improving the robustness of dual-processor cross-upgrades, the method comprising the following steps: S100: The host computer sends upgrade data to the slave DSP, which has no network interface, via the main DSP. The main DSP and the slave DSP exchange upgrade data through the dual-port RAM inside the FPGA and handshake each packet through the upgrade handshake register group. S200, the host computer sends an upgrade command, the main DSP writes a predetermined magic number to the upgrade trigger register and resets the power supply of the slave DSP, the slave DSP enters the upgrade mode after powering on and recognizing the predetermined magic number. S300 and main DSP write upgrade data into dual-port RAM in fixed-length data blocks, and simultaneously write the current package number into the package number register and set the write completion flag. S400: Align the local packet number with the packet number register value from the DSP, then read the data from the dual-port RAM and copy it to the readback verification RAM, and then set the read completion flag. The S500 and main DSP compare the readback verification RAM with the local cache. If they are inconsistent, the write completion flag is set to an error value and a retransmission is triggered. If they are consistent, the write completion flag is set and the verification is deemed successful. S600: After the data to be verified by the DSP is passed, the data is written to the NOR Flash and read back word by word for comparison. If it fails, the error writing flag is set and the writing is stopped. If they match, the completion flag is set and the main DSP is notified. S700: The main DSP reads the error flag and reports the upgrade result to the host computer. After all is completed, the main DSP clears the upgrade trigger register, resets the DSP, and jumps to the application area.

[0018] This invention eliminates the need for DSP network hardware and protocol stacks through an FPGA dual-port RAM intermediary architecture, significantly reducing hardware costs. Remote upgrade triggering from the DSP without a network is achieved through magic number triggering and power reset, requiring no on-site intervention. Packet number register alignment enables breakpoint resumption and retransmission alignment, eliminating the need to re-transmit after communication interruption recovery. A three-layer mechanism—packet number alignment, dual-port RAM readback verification, and NOR Flash word-by-word readback verification—works collaboratively to detect, retransmit, or safely terminate communication errors and write failures. Automatic reset and retry in case of DSP upgrade anomalies, combined with pre-erasure and word-by-word readback, prevents startup anomalies caused by program area remnants. The overall solution offers significant advantages in hardware cost, software simplicity, upgrade reliability, and maintainability.

[0019] To further understand the technical principles and effects of this technical solution, the following examples and comparative examples are used for illustration.

[0020] Example 1 Example 1 uses the above-described upgrade method to upgrade data, and its platform is as follows: The main DSP is a TI TMS320C6748, running the SYS / BIOS real-time operating system and equipped with an Ethernet PHY / MAC; The DSP is a TI TMS320C6748 (DSP2), running on bare metal with no network interface; The FPGA is a Xilinx Spartan-6; The NOR Flash is Micron M29EW, with 32M×16bit storage space; The host computer's fixed IP address is 192.168.1.2:80.

[0021] EMIF Address Mapping: Both the master DSP and the slave DSP access the FPGA through the EMIFA asynchronous bus. The FPGA base address is 0x62000000. Among them, the CS3 space 0x62000000-0x63FFFFFF maps the FPGA registers and dual-port RAM, and the CS2 space 0x60000000-0x61FFFFFF maps the NOR Flash program area.

[0022] Example of upgrading handshake register group (all based on FPGA module reset register base address 0x62530000): Upgrade trigger register DSP1TODSP2_GOTO_PROGARAM is located at offset 0x10 (FPGA_BASE+0x530000+0x10), write 0xB00A to notify the DSP to enter upgrade mode; read completion flag DSP2_RDDONE is located at offset 0x12; write completion flag DSP1_WRDONE is located at offset 0x14, 1 indicates verification passed, 2 indicates retransmission is required; error flag DSP2_PROGARAM_ERROR is located at offset 0x16, 0x00EE indicates NOR Flash programming failure; current packet number register DSP1_ONCEADDR is located at offset 0x18; total packet count register DSP1_TOTALLEN is located at offset 0x1C.

[0023] Dual-port RAM and readback RAM addresses: The dual-port RAM data buffer DSP1_BOOTRAM_ADDR is 0x62560000 (FPGA_BASE+0x560000), and 64 16-bit words are written per packet; the readback verification RAM base address is 0x62570000, and the same packet of data is written back to this area from the DSP for comparison by the master DSP.

[0024] The main DSP side process (corresponding to the hrt_uart_Progaram function): When the upgrade target is identified as DSP2, first write DSP1TODSP2_GOTO_PROGARAM=0xB00A, call Hrt_Co_PwrUP() to reset the slave DSP power supply, and then write DSP1_TOTALLEN=total number of packets; then enter the packet-by-packet loop: write DSP1_ONCEADDR=j, copy 64 16-bit words from the local cache ddrBuf+64×j to DSP1_BOOTRAM_ADDR, and set DSP1_WRDONE=1; wait for DSP2_RDDONE=1 and then clear it, read out the back-read verification RAM word by word and compare it with ddrBuf. If they are inconsistent, set DSP1_WRDONE=2. If the accumulated errors exceed 5, exit. If they are consistent, set DSP1_WRDONE=1; wait for DSP2_RDDONE=1 again and then clear it, read DSP2_PROGARAM_ERROR. If it is 0x00EE, return the failure code 0x63 to the host computer.

[0025] From the DSP side (corresponding to the DSP2 main function): After initializing EMIFA and GPIO, read DSP1TODSP2_GOTO_PROGARAM. If it is 0xB00A, enter the upgrade loop; first erase NOR... The first 15 blocks of Flash are processed, and the total number of packets (times) is read from DSP1_TOTALLEN. The loop continues from j=0 to times-1: after waiting for DSP1_WRDONE=1, it is cleared, and DSP1_ONCEADDR is read into num. If num≠j, j=num. 64 16-bit words are read from DSP1_BOOTRAM_ADDR into pBuf and written to the readback check area. DSP2_RDDONE=1 is set. DSP1_WRDONE is not 0. If it is 2, it is cleared and continues to wait for retransmission. If it is 1, it is cleared. The nor_flash_program is called word by word to write pBuf into M29EW_ADRS+j×128. Then, word by word is read back and compared. If they do not match, DSP2_PROGARAM_ERROR=0x00EE is written and the process exits. If they match, DSP2_RDDONE=1 is set.

[0026] Example 2 In Example 2, the communication link is interrupted in packet 5. The host computer retransmits data starting from packet 5. The main DSP writes packet number register = 5, reads num = 5 from the DSP and aligns it with local j. There is no need to repeatedly erase or re-burn the first 5 packets; it can directly continue from the breakpoint. Example 2 verifies the effectiveness of the packet number alignment mechanism in link jitter scenarios.

[0027] Example 3 In Example 3, the main DSP's TCP task connects to the host computer at 192.168.1.2:80 in client mode. After recving 522 bytes per frame, it calls `hrt_uart_copyto_fifo`, sharing the same FIFO packet parser with the UART received data. UART and network upgrades reuse the same upgrade scheduling function `hrt_uart_Program` on the main DSP side. Example 3 verifies the scalability of the upgrade channel.

[0028] Comparative Example 1 Comparative Example 1 is based on the same hardware platform as Example 1, but the readback verification RAM and corresponding verification steps are omitted. After the main DSP finishes writing the data block, it directly sets the write completion flag and directly writes the Flash after reading from the DSP.

[0029] Comparing Example 1 with Comparative Example 1, when a flip occurs in the dual-port RAM transmission channel, Comparative Example 1 fails to detect transmission errors, and erroneous data is directly burned into the Flash. Example 1, however, detects inconsistencies through readback verification and triggers retransmission, thus preventing erroneous writing to the Flash. Therefore, based on the data from Example 1 and Comparative Example 1, it can be seen that adding a readback verification step can achieve a 100% interception rate of transmission errors before burning.

[0030] Comparative Example 2 Comparative Example 2 is performed on the same hardware platform as Example 1, but the DSP does not use packet number register alignment and relies solely on local j-count incrementing. In Comparative Example 2, when the master DSP retransmits from packet 5 after a communication interruption, the slave DSP's local j remains at the value before the interruption (e.g., packet 3), resulting in packet number misalignment. Data from packet 5 is written to the Flash address of packet 3, causing data corruption in the program area. In contrast, Example 1, through packet number register alignment in step S400, automatically updates j to 5 in the slave DSP, ensuring correct address writing.

[0031] Comparative Example 3 Comparative Example 3 is based on the same hardware platform as Example 1. It uses a one-time burning method after the entire file TCP transmission is completed. The link is interrupted when the transmission reaches the 8th packet, and it needs to be retransmitted from the 0th packet. In contrast, Example 1 uses a breakpoint resume method. After the interruption of the 5th packet, it only needs to transmit from the 5th to the end. Assuming the total number of packets is 20, the retransmission amount is reduced from 20 packets to 15 packets, and the efficiency is improved by 25%.

[0032] Comparative Example 4 Comparative Example 4 is based on the same hardware platform as Example 1, but with an additional Ethernet PHY / MAC configured on the DSP and a lightweight TCP / IP protocol stack ported. The hardware cost increases by approximately 40 yuan (PHY / MAC chip and associated circuitry), the PCB area increases by approximately 15 cm², and the power consumption increases by approximately 200 mW. In Example 1, the DSP requires no network hardware, retaining only the NOR Flash and EMIF interfaces. Therefore, under the premise of equivalent upgrade functionality, Example 1 is superior to Comparative Example 4 in terms of cost, area, and power consumption.

[0033] In summary, the upgrade method provided by this invention utilizes packet number alignment to enable breakpoint resumption, ensuring that data is written to the correct address during retransmission; readback verification intercepts transmission errors before burning, ensuring that retransmitted data is correct; and Flash word-by-word readback detects write failures, ensuring the correct write result. The absence of any one of these three elements will affect the effectiveness of the other two. For example, without packet number alignment, data that passes readback verification may be written to the wrong address; without readback verification, packet number alignment only guarantees the correct address but the data may be incorrect. The synergistic effect of these three elements constitutes a complete closed loop for upgrade reliability.

[0034] The present invention has provided a detailed description of a reliable method for improving the robustness of dual-processor cross-upgrades. Specific examples have been used to illustrate the principles and implementation methods of the invention. The descriptions of these embodiments are merely illustrative of the core ideas of the invention. It should be noted that those skilled in the art can make various improvements and modifications to the invention without departing from its principles, and these improvements and modifications also fall within the scope of protection of the claims.

Claims

1. A reliable method for improving the robustness of dual-processor cross upgrade, characterized in that, The method includes the following steps: S100: The host computer sends upgrade data to the slave DSP, which has no network interface, via the main DSP. The main DSP and the slave DSP exchange upgrade data through the dual-port RAM inside the FPGA and handshake each packet through the upgrade handshake register group. S200, the host computer sends an upgrade command, the main DSP writes a predetermined magic number to the upgrade trigger register and resets the power supply of the slave DSP, the slave DSP enters the upgrade mode after powering on and recognizing the predetermined magic number. S300 and main DSP write upgrade data into dual-port RAM in fixed-length data blocks, and simultaneously write the current package number into the package number register and set the write completion flag. S400: Align the local packet number with the packet number register value from the DSP, then read the data from the dual-port RAM and copy it to the readback verification RAM, and then set the read completion flag. The S500 and main DSP compare the readback verification RAM with the local cache. If they are inconsistent, the write completion flag is set to an error value and a retransmission is triggered. If they are consistent, the write completion flag is set and the verification is deemed successful. S600: After the data to be verified by the DSP is passed, the data is written to the NOR Flash and read back word by word for comparison. If it fails, the error writing flag is set and the writing is stopped. If they match, the completion flag is set and the main DSP is notified. S700: The main DSP reads the error flag and reports the upgrade result to the host computer. After all is completed, the main DSP clears the upgrade trigger register, resets the DSP, and jumps to the application area.

2. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 1, characterized in that, The specific steps of step S100 are as follows: Construct a collaborative upgrade architecture of master DSP, slave DSP and FPGA. The host computer communicates with the master DSP via Ethernet TCP. The master DSP is connected to the FPGA via the EMIF asynchronous bus and exclusively occupies the remote communication resources. The slave DSP exchanges upgrade data with the master DSP through the dual-port RAM inside the same FPGA. The FPGA is divided into a dual-port RAM data buffer, a readback verification RAM and an upgrade handshake register. The upgrade handshake register includes an upgrade trigger register, a write completion flag register, a read completion flag register, a current packet number register, a total packet count register and an error flag register.

3. The reliable method for improving the robustness of dual-processor cross-upgrades as described in claim 2, characterized in that, The specific steps of step S200 are as follows: the host computer sends an upgrade command, the main DSP writes a predetermined magic number to the upgrade trigger register and performs a power reset operation on the slave DSP, the slave DSP reads the upgrade trigger register after power-on initialization, and enters the upgrade mode after recognizing the predetermined magic number; otherwise, it automatically resets and jumps to the application area through the watchdog timer. The predetermined magic number is 0xB00A.

4. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 3, characterized in that, Between steps S200 and 300, there is also step S200': after entering the upgrade mode from the DSP, the total number of packages is read from the total number of upgrade files, and all sectors of the program area in the NOR Flash are erased in advance.

5. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 2, characterized in that, The specific steps of step S300 are as follows: the main DSP writes the upgrade data into the dual-port RAM data buffer in fixed-length data blocks according to the package number order. After writing is completed, the current package number is written into the current package number register and the write completion flag register is set. The fixed-length data block is 64 16-bit words. The package number order is incremented from 0. After each package is written, the current package number register is written first and then the write completion flag is set.

6. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 2, characterized in that, The specific steps of step S400 are as follows: after the DSP polls until the write completion flag is set, it is cleared to zero. Then the current packet number register is read, and the local expected packet number is aligned with the value of the register. Then the packet number data is read from the dual-port RAM data buffer and copied to the readback verification RAM. The read completion flag register is then set.

7. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 6, characterized in that, The specific steps of step S500 are as follows: After the main DSP waits for the read completion flag to be set, it reads the data word by word from the readback verification RAM and compares it with the local cache. If they are inconsistent, the write completion flag is set to an error value. After the DSP detects the error value, it clears the flag and waits for the main DSP to resend the same packet. If they are consistent, the write completion flag is set and the verification is determined to be successful.

8. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 7, characterized in that, Specifically, step S600 involves: writing data word by word from the DSP into the NOR Flash; after writing is complete, reading back word by word and comparing it with the local cache; if any word is inconsistent, writing a predetermined error code to the error flag register and stopping the process. If a match is found, the read completion flag is set and the main DSP is notified to proceed to the next packet. The address of the NOR Flash is the program area base address plus the packet number multiplied by the packet length, and the predetermined error code is 0x00EE.

9. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 8, characterized in that, The specific steps of step S700 are as follows: After the main DSP detects the read completion flag, it clears it and reads the error flag register. If the predetermined error code 0x00EE is read, it reports the write failure to the host computer. Otherwise, it continues to send the next packet. After all data packets are sent, the main DSP reports the upgrade success to the host computer, clears the upgrade trigger register, resets the DSP, and jumps to the newly written application area.

10. The reliable method for improving the robustness of dual-processor cross-upgrade as described in claim 2, characterized in that, The readback verification RAM and the dual-port RAM data buffer are independently mapped regions in the FPGA address space.