A GPU data processing system based on big data

Abstract

The present invention relates to the field of GPU data processing technology, and in particular to a GPU data processing system based on big data, including a data preprocessing module, a reorganization module, a stream processor, a cache module, and a control module, which is used to determine the corresponding processing method according to the access frequency of standard graphic data or determine whether the rendering accuracy meets the requirements in combination with the proportion of the number of graphics frames with rendering distortion when the GPU data processing accuracy is determined to be unsatisfactory according to the average error amount of the displayed graphics; and if the rendering accuracy does not meet the requirements, the number of segmentation types of the rendering picture is adjusted or the data reorganization method of the reorganization module is adjusted according to the average graphic error change amount of several dynamic updates. The present invention improves the GPU data processing accuracy and the accuracy of graphics rendering.

Description

Technical Field

[0001] The present invention relates to the field of GPU data processing, and in particular to a GPU data processing system based on big data. Background Art

[0002] In the prior art, with the rapid rise of fields such as big data, artificial intelligence and deep learning, GPU data processing systems are currently in a stage of rapid development and widespread application. GPU has powerful parallel computing capabilities, adopts a multi-core architecture, and has multiple computing units or cores, enabling it to execute multiple tasks or threads simultaneously and has a high memory bandwidth. However, although GPU data processing systems have made significant progress, with the continuous growth of data volume and the increasing complexity of computing tasks, GPU data processing systems are facing the problem of low data processing precision and accuracy.

[0003] Chinese Patent Publication No.: CN105678681A discloses a GPU data processing method, GPU, PC architecture processor and GPU data processing system, including obtaining encoded data stored in a preset memory and hard decoding it into decoded data; storing the decoded data after hard decoding in a preset allocated video memory; rendering and displaying the decoded data extracted from the preset allocated video memory. It can be seen that the GPU data processing method, GPU, PC architecture processor and GPU data processing system have the problems of low GPU data processing accuracy and inaccurate graphics rendering. Summary of the invention

[0004] To this end, the present invention provides a GPU data processing system based on big data to overcome the problems of low GPU data processing accuracy and inaccurate graphics rendering in the prior art.

[0005] To achieve the above object, the present invention provides a GPU data processing system based on big data, comprising:

[0006] A data preprocessing module is used to preprocess the graphic source file to output standard graphic data;

[0007] a reorganization module, connected to the data preprocessing module, for classifying and reorganizing the standard graphic data to output a standard graphic data set;

[0008] a stream processor connected to the recombining module and used to generate display graphics according to the standard graphic data set, comprising a rendering unit for rendering the display graphics and a display update unit connected to the rendering unit for dynamically updating the display graphics;

[0009] A cache module, connected to the reorganization module, for caching the standard graphic data set and the display graphic;

[0010] A control module, which is respectively connected to the data preprocessing module, the reorganization module, the stream processor and the cache module, and is used to determine the corresponding processing method according to the access frequency of the standard graphic data or determine the rendering accuracy in combination with the proportion of the number of graphics frames with rendering distortion when it is determined that the GPU data processing accuracy does not meet the requirements according to the average error amount of the displayed graphics;

[0011] And, when the rendering accuracy does not meet the requirements, the number of segmentation types of the rendering picture is adjusted or the data reorganization mode of the reorganization module is adjusted according to the average graphic error variation of several dynamic updates;

[0012] The corresponding processing method includes adjusting the cache ratio of data or updating data from the log calling channel.

[0013] Furthermore, the data preprocessing module includes:

[0014] A data reading unit, used to read the graphic source file and load it to form basic graphic data;

[0015] A data cleaning unit connected to the data reading unit, for identifying errors, duplicates, missing values ​​and abnormal values ​​in the graphic basic data to form graphic data to be converted;

[0016] A standardization unit is connected to the data cleaning unit and is used to convert the format of the graphic data to be converted to form the standard graphic data.

[0017] Furthermore, it also includes:

[0018] A storage module, connected to the stream processor and the control module respectively, for storing display graphics formed by the stream processor;

[0019] A communication module is connected to the data preprocessing module, the reorganization module, the cache module and the control module respectively, and is used for data exchange with each module.

[0020] Furthermore, the control module is connected to the stream processor to obtain an average error amount of the display graphics, wherein if the average error amount is greater than a preset second error amount, the control module determines that the GPU data processing accuracy does not meet the requirements, and obtains the data access frequency to determine the corresponding processing method;

[0021] If the average error amount is greater than a preset first error amount and less than a preset second error amount, the control module preliminarily determines that the rendering accuracy does not meet the requirements.

[0022] Further, determining a corresponding processing method according to the access frequency of the standard graphic data includes determining the type of the standard graphic data according to a comparison result between the access frequency of the standard graphic data and a preset access frequency, and determining the corresponding processing method according to the type, wherein:

[0023] If the access frequency of the standard graphic data is greater than the preset access frequency, the control module determines that such standard graphic data is primary data; if the access frequency of the standard graphic data is less than the preset access frequency, the control module determines that such standard graphic data is secondary data;

[0024] Wherein, the types of the standard graphic data include primary data and secondary data;

[0025] If the access frequency of primary data is greater than that of secondary data, the data is updated from the log call channel; if the access frequency of primary data is less than that of secondary data, the cache ratio of standard graphic data is reduced;

[0026] Among them, the first-level data calls data updates from the log, and the second-level data calls data updates from the storage module.

[0027] Further, the control module is connected to the stream processor to obtain the proportion of the number of graphics frames with rendering distortion, and secondarily determine the rendering accuracy according to the comparison result of the proportion of the number of graphics frames and the preset second distortion proportion, and increase the number of segmentation types of the rendering picture according to the second determination result of the rendering accuracy, wherein,

[0028] If the proportion of the number of graphics frames with rendering distortion is greater than the preset second distortion proportion, the control module determines for the second time that the rendering accuracy does not meet the requirement, and increases the number of segmentation types of the rendering picture.

[0029] Furthermore, the calculation formula for the proportion of the number of graphics frames with rendering distortion is:

[0030]

[0031] Among them, N1 is the number of graphics frames with rendering distortion, and N2 is the total number of graphics frames formed after rendering.

[0032] Furthermore, the control module is connected to the stream processor to obtain the average graphic error variation of the several dynamic updates, and the uniformity of data reorganization is determined by combining the proportion of the number of graphic frames with rendering distortion and the average graphic error variation, and the data of the reorganization module is decomposed according to the determination result of the uniformity of the data reorganization, wherein:

[0033] If the proportion of the number of graphics frames with rendering distortion is greater than the preset first distortion proportion and less than the preset second distortion proportion, the control module preliminarily determines that the uniformity of data reconstruction does not meet the requirements; if the average graphic error change is greater than the preset graphic error change, the control module secondarily determines that the uniformity of data reconstruction does not meet the requirements and decomposes the data of the reconstruction module.

[0034] Furthermore, the extent of reduction in the data cache ratio is determined by a difference between the secondary data access frequency and the primary data access frequency.

[0035] Compared with the prior art, the beneficial effect of the present invention lies in that the processing system of the present invention realizes the cleaning and standardization of source file data by setting a data preprocessing module, so as to convert it into a format that can be understood and processed by the GPU, thereby improving the accuracy and consistency of the data, and thus improving the accuracy of subsequent processing and analysis; by setting a reorganization module to rearrange irregular data, efficient access and processing of data are realized, and the calculation process is further accelerated, while reducing irregular or indirect memory access, reducing the complexity of memory access, and improving the efficiency of data access to achieve performance improvement; by setting a control module, the average error of the displayed graphics is reduced, and the GPU data processing accuracy is improved, and the rendering accuracy is improved by reducing the number of graphics frames with rendering distortion, and the uniformity of data reorganization is improved by reducing the average change in the dynamic update of the graphics error, and further optimizes the data organization and layout of the system and improves the stability of the output results.

[0036] Furthermore, the big data-based GPU data processing system described in the present invention stores display graphics by setting up a storage module and a communication module, thereby increasing the processing speed of the GPU, and connects each module to directly transmit data, thereby improving the GPU data processing efficiency, optimizing data transmission and interaction processes, and improving the overall performance of the system.

[0037] Furthermore, the big data-based GPU data processing system of the present invention determines that the GPU data processing accuracy does not meet the requirements when the average error amount is greater than a preset second error amount, and obtains the data access frequency to determine the corresponding processing method, thereby further reducing the impact of redundant data generated during the operation of the GPU on the output display graphics, improving operation efficiency, improving the utilization of storage space in the cache module, and reducing bandwidth occupancy.

[0038] Furthermore, the big data-based GPU data processing system of the present invention improves the accuracy and clarity of displayed graphics by re-determining that the rendering accuracy does not meet the requirements when the proportion of the number of graphics frames with rendering distortion is greater than the proportion of a preset second distortion number, and increases the number of segmentation types of the rendering picture, thereby further improving the rendering accuracy and the quality of the rendering effect.

[0039] Furthermore, the big data-based GPU data processing system of the present invention determines that the uniformity of data reconstruction does not meet the requirements when the average graphic error change is greater than the preset graphic error change, and decomposes the data of the reconstruction module, thereby reducing the impact caused by the irregularity of the graphic data itself. By further decomposing the data set, data processing is made more efficient, and the accuracy of data analysis is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The overall structural block diagram of the GPU data processing system based on big data according to an embodiment of the present invention;

[0041] Figure 2 This is a structural block diagram of a data preprocessing module of a GPU data processing system based on big data according to an embodiment of the present invention;

[0042] Figure 3 This is a structural block diagram of a data preprocessing module and a control module of a GPU data processing system based on big data according to an embodiment of the present invention;

[0043] Figure 4 The present invention is a logical flow chart of a GPU data processing system based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0046] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0047] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] See also Figure 1 , Figure 2 , Figure 3 as well as Figure 4 As shown, they are respectively an overall structural block diagram of a GPU data processing system based on big data, a structural block diagram of a data preprocessing module, a structural block diagram of a data preprocessing module and a control module, and a logic flow chart. The GPU data processing system based on big data of the present invention includes:

[0049] A data preprocessing module is used to preprocess the graphic source file to output standard graphic data;

[0050] a reorganization module, connected to the data preprocessing module, for classifying and reorganizing the standard graphic data to output a standard graphic data set;

[0051] a stream processor connected to the recombining module and used to generate display graphics according to the standard graphic data set, comprising a rendering unit for rendering the display graphics and a display update unit connected to the rendering unit for dynamically updating the display graphics;

[0052] A cache module, connected to the reorganization module, for caching the standard graphic data set and the display graphic;

[0053] A control module, which is respectively connected to the data preprocessing module, the reorganization module, the stream processor and the cache module, and is used to determine the corresponding processing method according to the access frequency of the standard graphic data or determine the rendering accuracy in combination with the proportion of the number of graphics frames with rendering distortion when it is determined that the GPU data processing accuracy does not meet the requirements according to the average error amount of the displayed graphics;

[0054] And, when the rendering accuracy does not meet the requirements, the number of segmentation types of the rendering picture is adjusted or the data reorganization mode of the reorganization module is adjusted according to the average graphic error variation of several dynamic updates;

[0055] The corresponding processing method includes adjusting the cache ratio of data or updating data from the log calling channel.

[0056] Specifically, the types of segmentation of the rendering picture include segmentation of the rendering picture according to equal areas, segmentation of the rendering picture according to the number of rendering points, and segmentation of the rendering picture according to the proportion of the shadow area of ​​the rendering picture.

[0057] Specifically, the status update result is a data set composed of coordinate data of each pixel point of the display graphics after each update.

[0058] Specifically, the construction process of the display graphics module is a conventional technical means well known to those skilled in the art, so its construction process is not described in detail.

[0059] Specifically, the difference between the standard graphic data set and the standard graphic data is that the standard graphic data set is a data set that has been classified and reorganized, and the standard graphic data set includes several sub-data sets, and the data type of each sub-data set is different.

[0060] In implementation, the processing system of the present invention realizes the cleaning and standardization of source file data by setting a data preprocessing module so as to convert it into a format that can be understood and processed by the GPU, thereby solving the problem of low processing efficiency caused by data redundancy and inconsistency, thereby improving the accuracy and consistency of data, thereby improving the accuracy of subsequent processing and analysis; by setting a reorganization module to rearrange irregular data, the disorder of many irregular data is solved and reorganized, efficient access and processing of data are realized, and the calculation process is further accelerated, while reducing irregular or indirect memory access, reducing the complexity of memory access, and improving the efficiency of data access to achieve performance improvement; by setting a control module, the average error of displayed graphics is reduced, the GPU data processing accuracy is improved, the rendering accuracy is improved by reducing the number of graphics frames with rendering distortion, and the uniformity of data reorganization is improved by reducing the average change in dynamic updated graphics errors, further optimizing the data organization and layout of the system and improving the stability of the output results.

[0061] Specifically, the state update of the display graph by the display update unit includes converting edge deletion operations into edge additions or performing multiple updates on vertex merging processing for multiple snapshots in different time windows.

[0062] Specifically, the data preprocessing module includes:

[0063] A data reading unit, used to read the graphic source file and load it to form basic graphic data;

[0064] A data cleaning unit connected to the data reading unit, for identifying errors, duplicates, missing values ​​and abnormal values ​​in the graphic basic data to form graphic data to be converted;

[0065] A standardization unit is connected to the data cleaning unit and is used to convert the format of the graphic data to be converted to form the standard graphic data.

[0066] Specifically, the graphic basic data includes the size information of the graphic, the color information of the graphic, and the outline information of the graphic.

[0067] Specifically, compared with the basic graphics data, the amount of redundant data and erroneous data in the graphics data to be converted is much smaller than the amount of redundant data and erroneous data in the basic graphics data.

[0068] Specifically, erroneous data includes incomplete data, imperfect data, and data that obviously deviates from the normal range, such as incorrect coordinates and light and shadow information.

[0069] Specifically, the format of the standard graphic data is changed compared to the graphic data to be converted, for example, one of the formats of JPEG, PNG, and BMP is converted into one of the formats of CSC and CSR.

[0070] Specifically, the GPU data processing system based on big data of the present invention further includes:

[0071] A storage module, connected to the stream processor and the control module respectively, for storing display graphics formed by the stream processor;

[0072] A communication module is connected to the data preprocessing module, the reorganization module, the cache module and the control module respectively, and is used for data exchange with each module.

[0073] Specifically, the big data-based GPU data processing system described in the present invention stores display graphics by setting up a storage module and a communication module, thereby improving the processing speed of the GPU, and connecting each module directly for data transmission, thereby improving the GPU data processing efficiency, optimizing data transmission and interaction processes, and improving the overall performance of the system.

[0074] Specifically, the control module is connected to the stream processor to obtain an average error amount of the display graphics, wherein if the average error amount is greater than a preset second error amount, the control module determines that the GPU data processing accuracy does not meet the requirements, and obtains the data access frequency to determine the corresponding processing method;

[0075] If the average error amount is greater than a preset first error amount and less than a preset second error amount, the control module preliminarily determines that the rendering accuracy does not meet the requirements.

[0076] Specifically, the amount of error in the displayed graphics is quantified by using image processing software to compare the pixel differences between the graphics rendered by the GPU and the reference image, and then taking the average value.

[0077] Specifically, determining the corresponding processing method according to the access frequency of the standard graphic data includes determining the type of the standard graphic data according to the comparison result between the access frequency of the standard graphic data and the preset access frequency, and determining the corresponding processing method according to the type, wherein:

[0078] If the access frequency of the standard graphic data is greater than the preset access frequency, the control module determines that such standard graphic data is primary data; if the access frequency of the standard graphic data is less than the preset access frequency, the control module determines that such standard graphic data is secondary data;

[0079] Wherein, the types of the standard graphic data include primary data and secondary data;

[0080] If the access frequency of primary data is greater than that of secondary data, the data is updated from the log call channel; if the access frequency of primary data is less than that of secondary data, the cache ratio of standard graphic data is reduced;

[0081] Among them, the first-level data calls data updates from the log, and the second-level data calls data updates from the storage module.

[0082] Specifically, the access frequency is detected by a trigger.

[0083] Specifically, access frequency refers to the number of times the GPU accesses data within a period of time. This frequency depends on multiple factors, including the complexity of graphics rendering, the amount of data, the capacity and bandwidth of video memory, etc.

[0084] Specifically, the big data-based GPU data processing system of the present invention determines that the GPU data processing accuracy does not meet the requirements when the average error amount is greater than a preset second error amount, and obtains the data access frequency to determine the corresponding processing method, thereby further reducing the impact of redundant data generated during the operation of the GPU on the output display graphics, improving operation efficiency, improving the utilization of storage space in the cache module, and reducing bandwidth occupancy.

[0085] In implementation, the preset first error amount generally ranges from [50, 80], and the preset second error amount generally ranges from [80, 100].

[0086] Preferably, the preferred embodiments for presetting the first error amount are 60, and the preferred embodiments for presetting the second error amount are 90.

[0087] In practice, the preset access frequency generally ranges from [80,000 times / second to 120,000 times / second].

[0088] Preferably, the preferred embodiment of the preset access frequency is 100,000 times / second.

[0089] In implementation, for every time the difference between the average error amount of the displayed graphics and the second error amount exceeds 5, the proportion of data updated from the log call channel is increased by 0.1 or the cache ratio of standard graphics data is reduced by 0.1. For example, the average error amount of the actual graphics is 10. If the access frequency of the primary data is greater than the access frequency of the secondary data, the proportion of data updated from the log call channel is increased by 0.2 from the current level. If the access frequency of the primary data is less than the access frequency of the secondary data, the cache ratio of the standard graphics data is reduced by 0.2 from the current level.

[0090] Specifically, the control module is connected to the stream processor to obtain the proportion of the number of graphics frames with rendering distortion, and secondarily determine the rendering accuracy according to the comparison result of the proportion of the number of graphics frames and the preset second distortion proportion, and increase the number of segmentation types of the rendering picture according to the second determination result of the rendering accuracy, wherein,

[0091] If the proportion of the number of graphics frames with rendering distortion is greater than the preset second distortion proportion, the control module determines for the second time that the rendering accuracy does not meet the requirement, and increases the number of segmentation types of the rendering picture.

[0092] Specifically, the calculation formula for the proportion of the number of graphics frames with rendering distortion is:

[0093]

[0094] Among them, N1 is the number of graphics frames with rendering distortion, and N2 is the total number of graphics frames formed after rendering.

[0095] Specifically, the big data-based GPU data processing system of the present invention realizes improvement in the precision and clarity of displayed graphics, and further realizes improvement in rendering accuracy and improvement in the quality of rendering effect, by causing the control module to re-determine that the rendering accuracy does not meet the requirements when the proportion of the number of graphics frames with rendering distortion is greater than the proportion of the preset second distortion number, and increases the number of segmentation types of the rendering picture.

[0096] In implementation, the preset first distortion ratio is 0.002, and the preset second distortion ratio is 0.005.

[0097] In implementation, when the difference between the proportion of graphics frames with rendering distortion and the proportion of the preset second distortion quantity exceeds 0.0005, the number of segmentation types of the rendering picture increases by 2.

[0098] Specifically, the control module is connected to the stream processor to obtain the average graphic error variation of the several dynamic updates, and the uniformity of data reorganization is determined by combining the proportion of the number of graphic frames with rendering distortion and the average graphic error variation, and the data of the reorganization module is decomposed according to the determination result of the uniformity of the data reorganization, wherein:

[0099] If the proportion of the number of graphics frames with rendering distortion is greater than the preset first distortion proportion and less than the preset second distortion proportion, the control module preliminarily determines that the uniformity of data reconstruction does not meet the requirements; if the average graphic error change is greater than the preset graphic error change, the control module secondarily determines that the uniformity of data reconstruction does not meet the requirements and decomposes the data of the reconstruction module.

[0100] Specifically, the data of the reorganization module is decomposed to increase the quantity of the standard graphic data set.

[0101] Specifically, the extent of reduction in the data cache ratio is determined by the difference between the secondary data access frequency and the primary data access frequency.

[0102] Specifically, the big data-based GPU data processing system of the present invention determines that the uniformity of data reconstruction does not meet the requirements when the average graphic error change is greater than the preset graphic error change, and decomposes the data of the reconstruction module, thereby reducing the impact caused by the irregularity of the graph data itself. By further decomposing the data set, data processing is made more efficient, and the accuracy of data analysis is further improved.

[0103] Specifically, average graphics errors include loss of graphic details, color distortion, jagged image edges, and texture deformation, misalignment, or stretching.

[0104] Specifically, the formula for the average graph error change is:

[0105]

[0106] Among them, s is the average graphic error change, A i is the number of graphic errors in the i-th comparison, a is the number of comparisons, and a is a natural number greater than 1.

[0107] In implementation, in implementation, the preset graphic error variation generally ranges from [10, 30].

[0108] Preferably, the preferred embodiment of the preset graphic error change amount is 20.

[0109] In implementation, every time the difference between the dynamically updated average graphic error change and the preset graphic error change exceeds 50, the data of the reorganization module is decomposed and the number of standard graphic data sets increases by 50N. For example, when the dynamically updated average graphic error change is 30, the number of standard graphic data sets increases to 200+50×2=300N.

[0110] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A GPU data processing system based on big data, characterized in that: include: A data preprocessing module is used to preprocess the graphic source file to output standard graphic data; a reorganization module, connected to the data preprocessing module, for classifying and reorganizing the standard graphic data to output a standard graphic data set; a stream processor connected to the recombining module and used to generate display graphics according to the standard graphic data set, comprising a rendering unit for rendering the display graphics and a display update unit connected to the rendering unit for dynamically updating the display graphics; A cache module, connected to the reorganization module, for caching the standard graphic data set and the display graphic; a control module, which is connected to the data preprocessing module, the reorganization module, the stream processor and the cache module respectively, and is used to determine the corresponding processing method according to the access frequency of the standard graphic data or determine the rendering accuracy in combination with the proportion of the number of graphics frames with rendering distortion when it is determined that the GPU data processing accuracy does not meet the requirements according to the average error amount of the displayed graphics; And, when the rendering accuracy does not meet the requirements, the number of segmentation types of the rendering picture is adjusted or the data reorganization mode of the reorganization module is adjusted according to the average graphic error variation of several dynamic updates; The control module is connected to the stream processor to obtain an average error amount of the display graphics, wherein if the average error amount is greater than a preset second error amount, the control module determines that the GPU data processing accuracy does not meet the requirements, and obtains the data access frequency to determine the corresponding processing method; If the average error amount is greater than a preset first error amount and less than a preset second error amount, the control module preliminarily determines that the rendering accuracy does not meet the requirements; Determining a corresponding processing method according to the access frequency of the standard graphic data includes determining the type of the standard graphic data according to a comparison result between the access frequency of the standard graphic data and a preset access frequency, and determining the corresponding processing method according to the type, wherein: If the access frequency of the standard graphic data is greater than the preset access frequency, the control module determines that such standard graphic data is primary data; if the access frequency of the standard graphic data is less than the preset access frequency, the control module determines that such standard graphic data is secondary data; Wherein, the types of the standard graphic data include primary data and secondary data; If the access frequency of primary data is greater than that of secondary data, the data is updated from the log call channel; if the access frequency of primary data is less than that of secondary data, the cache ratio of standard graphic data is reduced; Among them, the first-level data calls data updates from the log, and the second-level data calls data updates from the storage module.

2. The GPU data processing system based on big data according to claim 1, characterized in that: The data preprocessing module comprises: A data reading unit, used to read the graphic source file and load it to form basic graphic data; A data cleaning unit connected to the data reading unit, for identifying errors, duplicates, missing values ​​and abnormal values ​​in the graphic basic data to form graphic data to be converted; A standardization unit is connected to the data cleaning unit and is used to convert the format of the graphic data to be converted to form the standard graphic data.

3. The GPU data processing system based on big data according to claim 1, characterized in that: Also includes: A storage module, connected to the stream processor and the control module respectively, for storing display graphics formed by the stream processor; A communication module is connected to the data preprocessing module, the reorganization module, the cache module and the control module respectively, and is used for data exchange with each module.

4. The GPU data processing system based on big data according to claim 1, characterized in that: The control module is connected to the stream processor to obtain the proportion of the number of graphics frames with rendering distortion, and secondarily determine the rendering accuracy according to the comparison result of the proportion of the number of graphics frames and the preset second distortion proportion, and increase the number of segmentation types of the rendering picture according to the second determination result of the rendering accuracy, wherein, If the proportion of the number of graphics frames with rendering distortion is greater than the preset second distortion proportion, the control module determines for the second time that the rendering accuracy does not meet the requirement, and increases the number of segmentation types of the rendering picture.

5. The GPU data processing system based on big data according to claim 4, characterized in that: The calculation formula for the proportion of the number of graphics frames with rendering distortion is: Among them, n is the proportion of graphics frames with rendering distortion, N1 is the number of graphics frames with rendering distortion, and N2 is the total number of graphics frames formed after rendering.

6. The GPU data processing system based on big data according to claim 5, characterized in that: The control module is connected to the stream processor to obtain the average graphic error variation of the several dynamic updates, and determine the uniformity of data reorganization by combining the proportion of the number of graphic frames with rendering distortion and the average graphic error variation, and decompose the data of the reorganization module according to the determination result of the uniformity of the data reorganization, wherein: If the proportion of the number of graphics frames with rendering distortion is greater than the preset first distortion proportion and less than the preset second distortion proportion, the control module preliminarily determines that the uniformity of data reconstruction does not meet the requirements; if the average graphic error change is greater than the preset graphic error change, the control module secondarily determines that the uniformity of data reconstruction does not meet the requirements and decomposes the data of the reconstruction module.

7. The GPU data processing system based on big data according to claim 6, characterized in that: The data of the reorganization module is decomposed to increase the quantity of the standard graphic data set.

8. The GPU data processing system based on big data according to claim 7, characterized in that: The reduction range of the data cache ratio is determined by the difference between the secondary data access frequency and the primary data access frequency.

Images