Data processing chip based on heterogeneous computing architecture

Abstract

The invention discloses a data processing chip based on a heterogeneous computing architecture, and relates to the technical field. The data processing chip based on the heterogeneous computing architecture comprises a data processing unit for receiving and preprocessing data to be transmitted; the clustering compression engine is used for carrying out clustering compression processing on the preprocessed data to be transmitted; the modulation and coding control unit is used for carrying out modulation and coding processing on the data to be transmitted after clustering compression processing; the redundant coding unit is used for carrying out redundant coding processing on the modulated and coded data to be transmitted; and outputting the to-be-transmitted data subjected to redundant coding processing to a protocol stack and interface unit of external equipment according to a set communication protocol. Through collaborative design of a clustering compression engine in a chip and a modulation coding control unit, clustering compression aiming at data structure characteristics and modulation strategy matching in a dynamic channel state are realized; the problems of fixed compression and modulation modes and lack of cooperation in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data processing chip based on a heterogeneous computing architecture. Background Art

[0002] Against the backdrop of the ongoing development of intelligent communications, edge computing, and high-performance data processing, traditional single-computing architecture chips are increasingly facing challenges such as low computational efficiency, high task coupling, and uneven resource utilization in multi-tasking scenarios such as processing complex structured data, dynamic modulation control, and secure coding. In particular, applications such as multimodal data fusion, high-speed communication protocol adaptation, and adaptive coding strategy execution place higher demands on the task heterogeneity, scheduling flexibility, and real-time responsiveness of computing units. To address these challenges, data processing chips based on heterogeneous computing architectures are becoming a hot topic in research and industrial development.

[0003] Based on the above solution, it is found that the limitations of the existing technology include at least the following problems. In the existing technology, when transmitting and processing multi-source heterogeneous data, there are generally problems such as fragmented processing flow, mismatched structural responses, and poor transmission adaptability. On the one hand, the data to be transmitted does not undergo a unified and standardized pre-processing process before entering the chip processing path. As a result, the subsequent compression link is difficult to perform accurate classification based on the data body structure, such as format differences, timing characteristics, or noise structure, resulting in problems such as uneven compression ratio and compression redundancy imbalance. On the other hand, the modulation and coding strategies are mostly fixed configurations, making it difficult to flexibly match the modulation method and coding strategy according to the volume differences of the compressed data packets and the current channel status, which can easily lead to increased bit error rate or waste of transmission redundancy. At the same time, the injection of redundant information lacks targeted control, and it is difficult to adjust the redundancy strength according to the importance of the signal data and the channel reliability. As a result, it is difficult to ensure the end-to-end transmission integrity of the data packet in complex environments. Finally, the output interface lacks dynamic orchestration capabilities in data protocol processing and external device docking, affecting transmission efficiency and compatibility.

[0004] Therefore, when facing data processing requirements under multi-structure and complex channel conditions, the existing technology lacks a closed-loop coordination mechanism in the overall chip structure, making it difficult to achieve full-link intelligent adaptation processing from data access to transmission output, which seriously restricts the stability, accuracy and resource utilization efficiency during heterogeneous data transmission. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a data processing chip based on a heterogeneous computing architecture, which solves the problems of structural response imbalance, modulation coding mismatch and inaccurate redundancy control in the transmission and processing of heterogeneous data in the existing technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a data processing chip based on a heterogeneous computing architecture, comprising: a data processing unit for receiving and preprocessing data to be transmitted; a clustering compression engine for performing clustering compression processing on the preprocessed data to be transmitted; a modulation and coding control unit for performing modulation and coding processing on the data to be transmitted after clustering compression processing; a redundant coding unit for performing redundant coding processing on the modulated and coded data to be transmitted; and a protocol stack and interface unit for outputting the redundantly coded data to be transmitted to an external device according to a set communication protocol.

[0007] Furthermore, the specific steps of clustering and compressing the pre-processed data to be transmitted are as follows: dividing the data to be transmitted into multiple data clusters based on a preset clustering algorithm; extracting the mean, variance and density within the cluster for the data samples in each data cluster, and analyzing the target compression ratio of each data cluster; performing data compression processing on each data cluster based on the target compression ratio to obtain several compressed data packets to be transmitted.

[0008] Furthermore, the specific formula for calculating the target compression ratio of a data cluster is as follows: ;in, is the target compression ratio of a data cluster, is the intra-cluster variance of a data cluster, is the density adjustment factor stored in the database, is the cluster mean of a data cluster, is the intra-cluster density of a data cluster, The compression adjustment coefficient stored in the database.

[0009] Furthermore, the specific steps of performing modulation and coding processing on the data to be transmitted after cluster compression processing are as follows: for each data packet to be transmitted, the modulation and coding score is analyzed separately, and a judgment analysis is performed with several preset modulation and coding score intervals, and each modulation and coding score interval corresponds to a modulation and coding strategy; according to the preset modulation and coding score interval to which the modulation and coding score of each data packet to be transmitted belongs, the corresponding modulation and coding strategy is matched, and modulation and coding processing is performed to generate several modulation and coding signal data to be transmitted.

[0010] Furthermore, the specific steps for analyzing the modulation and coding score of each data packet to be transmitted are as follows: for each data packet to be transmitted, identify the data volume respectively; obtain the current transmission status data of the transmission channel, including the current signal-to-noise ratio, the current bit error rate, the current delay jitter amplitude and the current bandwidth occupancy rate; based on the data volume and the current transmission status data of the channel, analyze the modulation and coding score of each data packet to be transmitted.

[0011] Furthermore, the specific formula for calculating the modulation and coding score of a data packet to be transmitted is as follows: ;in, Score the modulation and coding of a data packet to be transmitted, is the data volume of a data packet to be transmitted, is the current signal-to-noise ratio of the transmission channel, is the signal-to-noise ratio adjustment coefficient stored in the database, is the current bit error rate of the transmission channel, is the error adjustment factor stored in the database, is the current delay jitter amplitude of the transmission channel, is the delay sensitivity adjustment coefficient stored in the database, is the transmission stability adjustment coefficient stored in the database, is the current bandwidth occupancy of the transmission channel, is the bandwidth occupancy adjustment factor stored in the database, The bandwidth usage adjustment coefficient stored in the database.

[0012] Furthermore, the specific steps of performing redundant coding processing on the modulated and coded data to be transmitted are as follows: for each modulated and coded signal data to be transmitted, the redundant requirement score is analyzed separately, and a judgment analysis is performed with several preset redundant requirement score intervals, and each redundant requirement score interval corresponds to a redundant coding strategy; according to the preset redundant requirement score interval to which the redundant requirement score of each modulated and coded signal data to be transmitted belongs, the corresponding redundant coding strategy is matched, and redundant coding processing is performed to generate a redundant coding signal data set.

[0013] Furthermore, the specific steps for analyzing the redundancy requirement score of each modulation and coding signal data to be transmitted are as follows: for each modulation and coding signal data to be transmitted, the data volume is identified respectively; the historical transmission fault tolerance data of the transmission channel is obtained, including the historical packet loss rate and the historical number of retransmissions for several historical transmission cycles; the historical transmission fault tolerance data of the transmission channel is comprehensively analyzed to obtain the historical transmission fault tolerance characteristics of the transmission channel, including the historical average packet loss rate and the historical average number of retransmissions; the current bit error rate of the transmission channel is read, and a comprehensive analysis is performed in combination with the historical transmission fault tolerance characteristics of the transmission channel and the data volume of each modulation and coding signal data to be transmitted to obtain the redundancy requirement score of each modulation and coding signal data to be transmitted.

[0014] Furthermore, the specific formula for calculating the redundancy requirement score of a certain modulation and coding signal data to be transmitted is as follows: ;in, Score the redundancy requirement of a modulated coded signal data to be transmitted, is a natural constant, is the historical average packet loss rate of the transmission channel, is the packet loss impact adjustment coefficient stored in the database, is the historical average number of retransmissions of the transmission channel, is the current bit error rate of the transmission channel, is the retransmission error coupling adjustment coefficient stored in the database, is the data volume of a modulated coded signal data to be transmitted, is the volume perception adjustment coefficient stored in the database.

[0015] Furthermore, it also includes: a dynamic encryption and data protection unit, which is used to encrypt and verify the integrity of the data to be transmitted after cluster compression processing; and a storage and memory management unit, which is used to store and manage the data to be transmitted after redundant coding processing.

[0016] The present invention has the following beneficial effects: (1) The data processing chip based on heterogeneous computing architecture can perform cluster compression on the data to be transmitted after data preprocessing is completed by integrating a cluster compression engine and a modulation and coding control unit in the chip. Each data cluster determines the target compression ratio based on the extraction of the mean, variance and density characteristics within the cluster, thereby realizing differentiated compression processing for data structure differences. On this basis, the chip calculates the modulation and coding score according to the volume of each compressed data packet and the current channel status, including signal-to-noise ratio, bit error rate, delay jitter, and bandwidth occupancy, and matches the modulation and coding strategy for processing, ensuring that the compression strategy and modulation method achieve efficient coordination between data volume and channel characteristics. This solution avoids the problem in the existing technology that fixed compression and modulation methods are difficult to take into account the diverse channel status and data structure changes, significantly improves data compression efficiency and modulation matching flexibility, and enhances the adaptability of chip-level processing in dynamic communication scenarios.

[0017] (2) The data processing chip based on the heterogeneous computing architecture has an independent redundant coding unit after completing the modulation and coding processing, which is dedicated to analyzing the redundant demand score of each modulated and coded signal data, and combining the historical average packet loss rate, the historical average number of retransmissions and the current bit error rate of the channel to conduct a comprehensive evaluation. According to the interval of the redundant demand score, the corresponding redundant coding strategy is selected to inject redundant check fields into the data. In this way, the chip can adaptively adjust the redundancy strength, thereby improving the fault tolerance of data transmission under complex channel conditions and reducing transmission failure and retransmission overhead. Compared with the traditional fixed redundancy method, this solution can achieve more targeted redundant processing based on the volume difference of the data itself and the transmission history status. Under the premise of ensuring data integrity, it can effectively control the resource waste caused by redundant information, thereby achieving better transmission stability in the communication link.

[0018] (3) The data processing chip based on the heterogeneous computing architecture builds a complete heterogeneous computing task flow path, covering the processing processes such as data reception, compression, modulation, redundancy, encryption and protocol output, and is divided into independent units to ensure that each stage has controllable and adjustable parameter input and output. In particular, between cluster compression, modulation coding and redundancy processing, a scoring and interval matching mechanism is set up, so that the processing logic of each link can dynamically select a solution according to the upstream results, forming a traceable and tunable computing closed loop. This design effectively avoids the problems of poor module reusability and inability to track after process interruption in the existing technology due to tight coupling of processing logic, improves the processing integrity and response flexibility of the chip under multi-scenario adaptation, and has good scalability and module pluggable capabilities, which facilitates the rapid deployment of the chip in different application scenarios.

[0019] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a diagram of the data processing chip architecture based on a heterogeneous computing architecture of the present invention. DETAILED DESCRIPTION

[0021] See also Figure 1 An embodiment of the present invention provides a technical solution: a data processing chip based on a heterogeneous computing architecture, comprising: a data processing unit for receiving and preprocessing data to be transmitted; a clustering compression engine for performing clustering compression processing on the preprocessed data to be transmitted; a modulation and coding control unit for performing modulation and coding processing on the data to be transmitted after clustering compression processing; a redundant coding unit for performing redundant coding processing on the modulated and coded data to be transmitted; and a protocol stack and interface unit for outputting the redundantly coded data to be transmitted to an external device according to a set communication protocol.

[0022] The specific steps for preprocessing the received data to be transmitted are as follows: Denoising: Filter out the noise in the data using filtering algorithms such as low-pass filters and Kalman filters to ensure the purity of the input data.

[0023] Standardization: Converting data into a unified standard format or scale, such as Z-score standardization (mean is 0 and standard deviation is 1).

[0024] Data formatting: Ensure that data from different sources (such as sensor data, image data, etc.) are converted into a unified data structure for subsequent unit processing.

[0025] In terms of implementing the clustering compression engine, to ensure real-time response performance in high-frequency data transmission scenarios, the present invention adopts an implementation method that combines dedicated logic circuits with a parallel pipeline architecture. Specifically, the clustering operation module is constructed using a configurable array logic unit (such as an FPGA or ASIC), supporting instruction-level parallel execution of multiple clustering algorithms such as K-means and spectral clustering. The data preprocessing, distance calculation, and intra-cluster statistical processing units are designed as a multi-stage pipeline, corresponding to subtasks such as data reading, cluster identifier allocation, and compression ratio evaluation, respectively. This can significantly reduce the response delay of each compression round. In addition, after each cluster division, the internal cache area of ​​the compression engine adopts a double-ended queue structure to support asynchronous reading and compression writing, further improving hardware parallelism and transmission efficiency.

[0026] In physical implementation, this chip architecture is manufactured based on the 7nm FinFET process node. To meet the low-power operation requirements of edge computing devices, the present invention introduces a dynamic voltage and frequency scaling (DVFS) mechanism. This mechanism detects the task load and cache occupancy of each functional unit in real time and adjusts the supply voltage and clock frequency on demand. At the same time, a power-aware scheduling mechanism is adopted to automatically shut down redundant computing paths in non-high-concurrency scenarios to achieve dynamic power consumption optimization.

[0027] This invention is applicable to multiple input types, including structured and unstructured data. In unstructured data scenarios (such as images and videos), the data processing unit supports access to image vectors processed by visual encoders (such as CNN and ViT). The clustering compression engine provides adaptive clustering capabilities for high-dimensional continuous feature spaces, supports similarity clustering in high-dimensional feature maps, and combines multi-channel perceptual compression strategies to achieve effective compression of feature dimensions such as edges, textures, and colors in image data. The modulation strategy can be adaptively adjusted in combination with the image structure complexity index to ensure high-fidelity transmission.

[0028] The protocol stack and interface unit are designed based on a modular protocol adapter structure. They natively support mainstream protocols such as TCP / IP, UDP, HTTP, MQTT, and provide a custom protocol registration interface. They can quickly load external protocol description templates through register configuration and complete protocol adaptation and handshake during the communication initialization phase. The adaptation scope covers industrial automation, edge control, in-vehicle communication and low-power IoT protocol families.

[0029] Historical transmission data is maintained by a fault-tolerant data management unit and supports a periodic rolling update mechanism: after each transmission cycle, the transmission statistics and redundancy decision results of this round are appended to the database, and an elimination strategy is executed every 1,000 consecutive cycles or when the storage limit is exceeded. Only high-confidence data windows within the past 30 days are retained as the modeling basis for the scoring formula. For scenarios where channel status changes suddenly, the system introduces a rapid instability detection mechanism. If the scoring error deviates sharply for three consecutive rounds, the warning sign is automatically triggered and the system switches to the default redundancy level to ensure transmission robustness. In addition, each unit in the chip supports hot backup and failover mechanisms, including master / slave engine switching, cache mirror protection, and abnormal reinitialization processes.

[0030] This invention has completed engineering-level prototype chip verification, and measured performance under typical test conditions is as follows: compression ratio increased by 38.6% (compared to a non-clustered compression baseline), modulation and coding response latency was kept below 2.4ms, and redundant coding efficiency (fault recovery rate per bit of overhead) increased by approximately 21.3%. Testing was conducted simultaneously in the 5G sub-6GHz band and Wi-Fi 6 network environments. All parameters were averaged after 500 consecutive simulated transmission cycles and 20 actual comparison tests. Independent third-party testing will be conducted simultaneously.

[0031] Specifically, the specific steps of clustering and compressing the pre-processed data to be transmitted are as follows: Based on a preset clustering algorithm (such as K-means, spectral clustering, etc.), the data to be transmitted is divided into multiple data clusters. Specifically, the preset clustering algorithm is first called to perform structural division processing on all data samples. Each data to be transmitted contains multiple feature fields that have been normalized or standardized. The clustering compression engine uses methods such as K-means, spectral clustering, or density clustering based on the numerical distance or spatial similarity between these feature fields to automatically divide the entire data set into several data clusters with high internal similarity and low external similarity. Each data cluster contains several data records with similar structural features. After clustering is completed, the clustering compression engine assigns a unique cluster identifier to each data cluster, and stores the data samples in each cluster and the corresponding cluster number in a structured form in the cache area as input for subsequent feature extraction and compression decision operations.

[0032] For the data samples in each data cluster, the cluster mean, variance and density are extracted, and the target compression ratio of each data cluster is analyzed. Specifically, the cluster feature extraction operation is performed on the data samples in each data cluster in the cache, and the mean, variance and distribution density information of the samples in the cluster on the key feature dimensions are respectively counted. The mean is used to describe the concentrated position of the cluster samples on the main feature dimensions, the variance is used to characterize the fluctuation amplitude and discreteness between samples, and the density represents the concentration of the number of samples in the unit feature space within the cluster; based on the above three statistical features and combined with the set compression adjustment strategy, the cluster compression engine performs compression potential assessment and compression ratio calculation on each data cluster, and outputs the compression target parameter representing the proportion of the number of samples to be retained in each data cluster. The compression ratio result will serve as the basis for subsequent sample screening and compression merging processing.

[0033] Based on the target compression ratio, data compression processing is performed on each data cluster separately to obtain several compressed data packets to be transmitted. Specifically, the number of data samples to be retained in the current cluster is determined, which is calculated by multiplying the target compression ratio by the total number of original data samples, and rounding up the number of retained samples; then, based on the characteristic density distribution of the current data cluster, a density-aware optimization strategy is adopted to select the most representative data points from the original samples to form a compressed subset; then, a structural recoding operation is performed on the data samples in the compressed subset, unifying the sample format and appending the compression parameter information, cluster number information and statistical summary information of the cluster to generate a compressed result data unit with a complete field structure; finally, the chip encapsulates each compressed result data unit into a data packet in a standard format as the input object of the modulation and coding control unit. This data packet is the compressed data packet to be transmitted and enters the subsequent modulation and coding processing flow.

[0034] The specific formula for calculating the target compression ratio of a data cluster is as follows: ;in, is the target compression ratio of a data cluster, is the intra-cluster variance of a data cluster, It is the density adjustment factor stored in the database, used to prevent the denominator from being 0. is the cluster mean of a data cluster, is the intra-cluster density of a data cluster, The compression adjustment coefficient stored in the database.

[0035] It should be explained that the compression adjustment coefficient stored in the database The specific acquisition steps are as follows: first, in the historical data compression operation, the original records of the intra-cluster variance, intra-cluster mean, intra-cluster density and the corresponding proportion of compressed samples of multiple compressed data clusters are recorded; then, a compression behavior mapping model is constructed based on these historical data, and the adjustment parameters that make the compression ratio prediction value closest to the actual compression result are extracted through the minimum mean square error fitting method; in the training process, the variance-to-density ratio and the intra-cluster mean are used as input variables, and the actual compression ratio is used as the target output, and iterative training is performed, and the change of the loss function is evaluated after each round of iteration; finally, when the compression prediction error tends to be stable and lower than the set error threshold in several consecutive rounds of iterations, the corresponding adjustment coefficient value in the calculation of this round is solidified as The current version of .

[0036] In this implementation, a cluster-driven compression mechanism is introduced so that data compression operations are no longer executed with a global unified strategy, but are customized based on the local characteristics of data distribution. Specifically, the chip uses a clustering compression engine to divide the preprocessed data into several clusters according to feature similarity, and then extracts and analyzes the mean, variance, and density within each cluster. The target compression ratio is calculated based on the set adjustment factor, and then representative samples are selected proportionally to generate a structurally complete compressed data packet. This mechanism significantly improves the accuracy and adaptability of the compression processing. It can not only effectively retain the core information of the data within the cluster, but also avoid the compression distortion problem caused by traditional compression methods that ignore data heterogeneity. At the same time, the compression adjustment coefficient is obtained by dynamic training of historical data, which makes the compression behavior have data-driven adaptive capabilities. The overall design realizes the differentiation, intelligence, and structural fidelity of the data compression process, providing higher-quality input data for subsequent modulation coding and transmission strategies, and enhancing the chip's processing efficiency and data integrity assurance capabilities in various communication tasks.

[0037] Specifically, the specific steps of performing modulation and coding processing on the data to be transmitted after cluster compression processing are as follows: for each data packet to be transmitted, the modulation and coding score is analyzed separately, and a judgment analysis is performed with several preset modulation and coding score intervals, and each modulation and coding score interval corresponds to a modulation and coding strategy; according to the preset modulation and coding score interval to which the modulation and coding score of each data packet to be transmitted belongs, the corresponding modulation and coding strategy is matched, and modulation and coding processing is performed to generate several modulation and coding signal data to be transmitted.

[0038] Modulation and coding of the data to be transmitted after cluster compression processing includes but is not limited to the following examples: The modulation and coding control unit receives three data packets to be transmitted after cluster compression processing, which are recorded as data packet A, data packet B and data packet C. After compression and encapsulation, each data packet contains the corresponding data volume field, compression ratio field and cluster number field.

[0039] Before performing modulation and coding processing, the modulation and coding control unit obtains the transmission status data of the current channel in real time, including: the signal-to-noise ratio is 18.7dB, the bit error rate is 2.6×10 -3 The delay jitter amplitude was 3.2ms, and the bandwidth utilization rate was 78%. Subsequently, the modulation and coding control unit calculated the modulation and coding score of each data packet based on the data volume and current channel state parameters, and obtained the following scores: packet A scored 0.78, packet B scored 0.49, and packet C scored 0.23.

[0040] The preset modulation and coding scoring intervals are divided into three categories: Modulation and coding score ≥ 0.70: 64QAM modulation + LDPC coding is used; Modulation and coding score values ​​range from 0.40 to 0.70: 16QAM modulation + convolutional code is used; Modulation and coding score < 0.40: QPSK modulation + BCH coding is used.

[0041] According to the score value, the modulation and coding control unit matches data packet A to the first strategy interval and selects the 64QAM+LDPC combination; matches data packet B to the second interval and adopts the 16QAM+convolutional code combination; matches data packet C to the third interval and executes the QPSK+BCH combination.

[0042] Finally, the modulation and coding control unit completes the modulation and coding processing for each data packet in turn, generates a corresponding modulation and coding signal data set, and transmits it to the redundant coding unit for subsequent processing.

[0043] The specific steps for analyzing the modulation and coding score of each data packet to be transmitted are as follows: for each data packet to be transmitted, identify the data volume; obtain the current transmission status data of the transmission channel, including the current signal-to-noise ratio, current bit error rate, current delay jitter amplitude and current bandwidth occupancy; based on the data volume and the current transmission status data of the channel, analyze the modulation and coding score of each data packet to be transmitted.

[0044] The specific formula for calculating the modulation and coding score of a data packet to be transmitted is as follows: ;in, Score the modulation and coding of a data packet to be transmitted, is the data volume of a data packet to be transmitted, is the current signal-to-noise ratio of the transmission channel, is the signal-to-noise ratio adjustment coefficient stored in the database, is the current bit error rate of the transmission channel, It is the error adjustment factor stored in the database, used to prevent the denominator from being 0. is the current delay jitter amplitude of the transmission channel, is the delay sensitivity adjustment coefficient stored in the database, is the transmission stability adjustment coefficient stored in the database, is the current bandwidth occupancy of the transmission channel, The bandwidth usage adjustment factor stored in the database is used to prevent the denominator from being 0. The bandwidth usage adjustment coefficient stored in the database.

[0045] It should be explained that the signal-to-noise ratio adjustment coefficient stored in the database , delay sensitivity adjustment coefficient , transmission stability adjustment coefficient , bandwidth occupancy adjustment coefficient The specific acquisition steps are as follows: calling several historical transmission records, which include the signal-to-noise ratio value, delay jitter amplitude, bit error rate, bandwidth occupancy rate during data packet transmission, as well as the modulation and coding strategy adopted under the corresponding conditions and the data packet transmission success rate under the strategy; then, based on the above historical data, a multivariate regression modeling method is used to construct a response function between the modulation and coding score value and each channel parameter, and the influence weight of each channel parameter on the score change is extracted by the minimum error fitting method; then, each weight is normalized and solidified into the corresponding 、 、 、 , and stored in the database for subsequent scoring function calls; among them, the signal-to-noise ratio adjustment coefficient Characterizes the sensitivity of the signal-to-noise ratio to score improvement, delay sensitivity adjustment coefficient Characterizes the attenuation response rate of the score to jitter fluctuations and the transmission stability adjustment coefficient Indicates the increase in the lower limit of the score in an unstable link, and the bandwidth occupancy adjustment coefficient Used to indicate the linear enhancement strength of bandwidth usage on the score penalty term.

[0046] In this implementation, a modulation and coding scoring mechanism is introduced to enable the modulation and coding control process to have real-time self-adaptation capabilities. After obtaining the data volume of each compressed data packet and the current channel status (such as signal-to-noise ratio, bit error rate, delay jitter and bandwidth occupancy), the chip uses a pre-trained scoring function to perform scoring calculations on each data packet, and accurately matches the applicable modulation and coding strategy combination based on the scoring results. This mechanism ensures that high-quality channel resources are preferentially matched with high-order modulation and high-efficiency coding, while low-quality channels automatically adopt more robust solutions to maximize the data transmission success rate and resource utilization efficiency. In addition, the various adjustment coefficients in the scoring function are derived from historical data modeling, which can dynamically reflect the actual impact of various channel parameters on transmission performance, and have stability and robustness. Compared with the traditional fixed strategy configuration method, this scoring mechanism greatly improves the chip's strategy selection accuracy and processing flexibility in complex transmission environments, effectively taking into account transmission reliability and efficiency, and is the key foundation for realizing intelligent coding control.

[0047] Specifically, the specific steps of performing redundant coding processing on the modulated and coded data to be transmitted are as follows: for each modulated and coded signal data to be transmitted, the redundant requirement score is analyzed separately, and a judgment analysis is performed with several preset redundant requirement score intervals, and each redundant requirement score interval corresponds to a redundant coding strategy; according to the preset redundant requirement score interval to which the redundant requirement score of each modulated and coded signal data to be transmitted belongs, the corresponding redundant coding strategy is matched, and redundant coding processing is performed to generate a redundant coded signal data set, including several redundant coded signal data to be transmitted.

[0048] The redundant coding processing of the modulated and coded data to be transmitted includes, but is not limited to, the following example: the redundant coding unit receives three modulated and coded signal data, respectively recorded as signal data A, signal data B, and signal data C. Each signal data comes from a clustered and compressed data packet, has completed the modulation and coding processing, and includes information such as an actual code volume field, a modulation mode field, a channel number, and a generation timestamp.

[0049] The redundant coding unit first obtains the following status parameters based on the channel fault tolerance status acquisition module in the current transmission cycle: the historical average packet loss rate is 3.2%, the bit error rate is 2.6×10 -3 The historical average number of retransmissions is 1.4. Subsequently, the redundant coding unit calculates the redundancy requirement score for each signal data based on the physical characteristics of each signal data (the coding volume is 3.6KB, 2.2KB, and 5.1KB, respectively). The scoring function uses a preset formula in the database, resulting in the following score: Signal data A: Redundancy requirement score = 0.61; Signal data B: Redundancy requirement score = 0.42; Signal data C: Redundancy demand score = 0.78; The preset redundancy requirement scoring ranges are as follows: Redundancy requirement score ≥ 0.70: Low-density parity check (LDPC) coding is used, and the redundancy ratio is set to 1:1; Redundancy requirement score between 0.40 and 0.70: Use convolutional code with a redundancy ratio of 3:2. Redundancy requirement score < 0.40: Use BCH encoding and set the redundancy ratio to 5:4.

[0050] Match results based on rating: Signal data A: processed using convolutional code; Signal data B: processed using convolutional code; Signal data C: LDPC coding is used.

[0051] The redundant coding unit implements the selected strategy, encoding each signal data into a data block with a corresponding redundant structure, injecting a redundancy check field, and generating three redundant coded signal data. Ultimately, these redundant coded signal data are combined to form a redundant coded signal data set, which serves as the input for the subsequent protocol encapsulation module.

[0052] The specific steps for analyzing the redundancy requirement score of each modulation and coding signal data to be transmitted are as follows: for each modulation and coding signal data to be transmitted, identify the data volume respectively; obtain the historical transmission fault tolerance data of the transmission channel, including the historical packet loss rate and the historical number of retransmissions for several historical transmission cycles; perform a comprehensive analysis (i.e., mean analysis) on the historical transmission fault tolerance data of the transmission channel to obtain the historical transmission fault tolerance characteristics of the transmission channel, including the historical average packet loss rate and the historical average number of retransmissions; read the current bit error rate of the transmission channel, and perform a comprehensive analysis based on the historical transmission fault tolerance characteristics of the transmission channel and the data volume of each modulation and coding signal data to be transmitted to obtain the redundancy requirement score of each modulation and coding signal data to be transmitted.

[0053] The specific formula for calculating the redundancy requirement score of a certain modulation and coding signal data to be transmitted is as follows: ;in, Score the redundancy requirement of a modulated coded signal data to be transmitted, is a natural constant, and in this embodiment, its value is 2.71. is the historical average packet loss rate of the transmission channel, is the packet loss impact adjustment coefficient stored in the database, is the historical average number of retransmissions of the transmission channel, is the current bit error rate of the transmission channel, is the retransmission error coupling adjustment coefficient stored in the database, is the data volume of a modulated coded signal data to be transmitted, is the volume perception adjustment coefficient stored in the database.

[0054] It should be explained that the packet loss impact adjustment coefficient stored in the database , retransmission error coupling adjustment coefficient , volume perception adjustment coefficient The specific acquisition steps are as follows: calling a data set of multiple communication periods from the historical record database, the data set includes: the packet loss rate, bit error rate, number of retransmissions, the coded volume, the actually adopted redundant coding strategy and the transmission success rate of each data packet under different channel conditions; then, using this batch of data as a training sample, constructing a response function between the redundant score value and the packet loss rate, the error retransmission combination factor and the coded volume, and iteratively optimizing each adjustment coefficient in the formula through a nonlinear multivariate minimum error fitting method, so that the score result and the actually selected redundant strategy can be kept consistent to the greatest extent; after the convergence condition is met, the adjustment coefficients obtained from this round of fitting are solidified as 、 、 .

[0055] This implementation scheme introduces a redundancy requirement scoring mechanism based on historical fault-tolerance status and data volume characteristics to achieve differentiated redundant processing strategy allocation for modulated coded signal data. Unlike the traditional unified redundant configuration method, this scheme combines the actual coding volume of each data to be transmitted with the packet loss rate, bit error rate, and number of retransmissions of the transmission channel in the past multiple cycles to dynamically calculate a more targeted scoring value, thereby automatically matching the optimal redundant coding strategy (such as LDPC, convolutional code, BCH) according to the scoring interval. This method can significantly improve the data transmission success rate in high-error channels, while effectively reducing the redundancy ratio in stable channels and saving transmission resources. The adjustment coefficient in the scoring function is extracted from historical communication data through nonlinear fitting optimization to ensure its high adaptability and accuracy to the current communication status. The overall mechanism has the advantages of scalability, trainability, and strong adaptability, providing strong support for data processing chips to achieve efficient and secure signal coding in complex communication scenarios.

[0056] Specifically, it also includes: a dynamic encryption and data protection unit, which is used to encrypt and integrity check the data to be transmitted after cluster compression processing, which is specifically as follows: first, the unit receives a set of data packets that have completed cluster compression, and each data packet contains fields such as data compression body, cluster number, compression ratio and processing timestamp; then, the unit determines the encryption level that should be adopted for the data packet based on the task identifier and data attributes, and calls the corresponding dynamic key from the database, and uses a symmetric encryption algorithm (such as AES-GCM or ChaCha20-Poly1305) to encrypt the data packet content to generate a ciphertext data block; then, the unit calculates the integrity summary value for each ciphertext data block, and uses a hash check algorithm (such as SHA-256 or HMAC) to perform a hash summary on the compressed body and metadata to generate a data integrity check code; finally, the unit encapsulates the ciphertext data and the corresponding check code as the input data structure of the subsequent modulation and coding control unit, and ensures that the structure has tamper-resistant and anti-replay capabilities in subsequent transmission.

[0057] The storage and memory management unit is used to store and manage the data to be transmitted after redundant coding processing. Specifically, the unit receives the redundant coding signal data set output by the redundant coding unit, and each data object includes a redundant field structure, original data index, redundant ratio and output sequence number; then, the unit determines the data storage location according to the current storage control strategy, and performs format standardization processing on the data object, including data block alignment, head and tail identifier supplementation and timestamp encoding; then, the unit sorts each data block according to the write priority and divides it into different cache areas or disk storage channels, and assigns it a unique storage identifier and retrieval index; finally, the unit records the current memory occupancy rate, block write status and data life cycle information in real time, and constructs a structured index table for data scheduling and read-write control during subsequent protocol encapsulation and external interface calls, ensuring the traceability and scheduling efficiency of redundant coding data in the chip's internal storage chip.

[0058] In this implementation plan, a dynamic encryption and data protection unit as well as a storage and memory management unit are introduced to provide data with stronger security protection and structured storage capabilities before transmission. The encryption protection part not only flexibly matches the encryption level based on data attributes, but also uses dynamic keys and high-strength symmetric encryption algorithms (such as AES-GCM, ChaCha20-Poly1305) to encrypt compressed data packets in real time. At the same time, it combines integrity check codes (such as SHA-256 or HMAC) to ensure the tamper-resistance and verifiability of data in subsequent processes, effectively preventing the risks of data replay and tampering. The storage management part performs standardized processing and priority scheduling based on the redundantly encoded data structure, and builds a storage mechanism that supports fast reading and writing and index retrieval. Through the storage identifier and lifecycle management mechanism, efficient scheduling and precise calling of data within the chip are achieved. Overall, this part of the design significantly improves the data processing chip's capabilities in data security, integrity protection and efficient resource management, and is suitable for communication processing scenarios with multi-tasking, high concurrency and high security requirements.

[0059] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0060] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A data processing chip based on a heterogeneous computing architecture, characterized in that: include: A data processing unit, used for receiving data to be transmitted and performing pre-processing; Cluster compression engine, used to perform cluster compression processing on the pre-processed data to be transmitted; A modulation and coding control unit, configured to perform modulation and coding processing on the data to be transmitted after cluster compression processing; A redundant coding unit, configured to perform redundant coding on the modulated and coded data to be transmitted; The protocol stack and interface unit are used to output the data to be transmitted after redundant coding processing to the external device according to the set communication protocol.

2. The data processing chip based on heterogeneous computing architecture according to claim 1, characterized in that: The specific steps for clustering and compressing the pre-processed data to be transmitted are as follows: Divide the data to be transmitted into multiple data clusters based on a preset clustering algorithm; For the data samples in each data cluster, extract the mean, variance and density within the cluster, and analyze the target compression ratio of each data cluster; Data compression is performed on each data cluster based on the target compression ratio to obtain a number of compressed data packets to be transmitted.

3. The data processing chip based on heterogeneous computing architecture according to claim 2, characterized in that: in, The specific formula for calculating the target compression ratio of a data cluster is as follows: ;in, 、 、 、 They are the target compression ratio, intra-cluster variance, intra-cluster mean, and intra-cluster density of a data cluster, respectively. 、 They are the density adjustment factor and the compression adjustment coefficient stored in the database respectively.

4. The data processing chip based on heterogeneous computing architecture according to claim 2, characterized in that: The specific steps of performing modulation and coding processing on the data to be transmitted after cluster compression processing are as follows: For each data packet to be transmitted, the modulation and coding score is analyzed separately, and a judgment analysis is performed with a number of preset modulation and coding score intervals, and each modulation and coding score interval corresponds to a modulation and coding strategy; According to the preset modulation and coding score interval to which the modulation and coding score of each data packet to be transmitted belongs, a corresponding modulation and coding strategy is matched, and modulation and coding processing is performed to generate a plurality of modulation and coding signal data to be transmitted.

5. The data processing chip based on heterogeneous computing architecture according to claim 1, characterized in that: The specific steps for analyzing the modulation and coding score of each data packet to be transmitted are as follows: For each data packet to be transmitted, identify the data volume; Obtain the current transmission status data of the transmission channel, including the current signal-to-noise ratio, current bit error rate, current delay jitter amplitude and current bandwidth occupancy; Based on the data volume and the current transmission status data of the channel, the modulation and coding score of each data packet to be transmitted is analyzed.

6. The data processing chip based on heterogeneous computing architecture according to claim 5, characterized in that: The specific formula for calculating the modulation and coding score of a data packet to be transmitted is as follows: ;in, 、 are the modulation and coding score and data volume of a data packet to be transmitted, respectively. 、 、 、 They are the current signal-to-noise ratio, current bit error rate, current delay jitter amplitude, and current bandwidth occupancy of the transmission channel, respectively. 、 、 、 、 、 The order is the signal-to-noise ratio adjustment coefficient, error adjustment factor, delay sensitivity adjustment coefficient, transmission stability adjustment factor, bandwidth occupancy adjustment factor, and bandwidth occupancy adjustment factor stored in the database.

7. The data processing chip based on heterogeneous computing architecture according to claim 5, characterized in that: The specific steps of performing redundant coding processing on the modulated and coded data to be transmitted are as follows: For each modulated coded signal data to be transmitted, the redundancy requirement score is analyzed separately, and a judgment analysis is performed with a plurality of preset redundancy requirement score intervals, and each redundancy requirement score interval corresponds to a redundant coding strategy; According to the preset redundancy requirement score interval to which the redundancy requirement score of each modulated coded signal data to be transmitted belongs, a corresponding redundant coding strategy is matched, and redundant coding processing is performed to generate a redundant coded signal data set.

8. The data processing chip based on heterogeneous computing architecture according to claim 7, characterized in that: The specific steps for analyzing the redundancy requirement score of each modulation and coding signal data to be transmitted are as follows: For each modulated coded signal data to be transmitted, respectively identifying the data volume; Obtain historical transmission fault tolerance data of the transmission channel, including historical packet loss rates and historical retransmission times for several historical transmission cycles; Comprehensively analyze the historical transmission fault tolerance data of the transmission channel to obtain the historical transmission fault tolerance characteristics of the transmission channel, including the historical average packet loss rate and the historical average number of retransmissions; The current bit error rate of the transmission channel is read, and a comprehensive analysis is performed based on the historical transmission fault tolerance characteristics of the transmission channel and the data volume of each modulation and coding signal data to be transmitted to obtain the redundancy requirement score of each modulation and coding signal data to be transmitted.

9. The data processing chip based on heterogeneous computing architecture according to claim 8, characterized in that: The specific formula for calculating the redundancy requirement score of a certain modulation and coding signal data to be transmitted is as follows: ； in, 、 The redundancy requirement score and data volume of a certain modulation and coding signal data to be transmitted are respectively is a natural constant, 、 、 They are the historical average packet loss rate, historical average number of retransmissions, and current bit error rate of the transmission channel, respectively. 、 、 They are the packet loss impact adjustment coefficient, retransmission error coupling adjustment coefficient, and volume perception adjustment coefficient stored in the database.

10. The data processing chip based on heterogeneous computing architecture according to claim 1, characterized in that: Also includes: Dynamic encryption and data protection unit, used to encrypt and verify the integrity of the data to be transmitted after cluster compression processing; The storage and memory management unit is used to store and manage the data to be transmitted after redundant coding processing.

Images