An intelligent management system and method applied to adaptive graphics rendering

Abstract

The application discloses an intelligent management system and method applied to adaptive graphic rendering, relates to the technical field of graphic rendering management, obtains a device running state label by analyzing a device running state, stores the device running state label into a historical database, divides a rendering intensity of a subtask rendering by considering the workload of a graphic rendering subtask, predicts a device running state after rendering of the graphic rendering subtask, plans the rendering intensity of the graphic rendering subtask, considers the rendering intensity of the graphic rendering subtask, selects an optimal rendering scheme, modifies an effective reference period of a database according to a rendering result after execution of the optimal rendering scheme is completed, and the application divides a reasonable rendering level range for each subtask, adopts a dynamic adjustment mode to make the rendering effect optimal within the device bearing capacity range, and takes into account the demand of a user for graphic quality and the device running efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of graphic rendering management, in particular to an intelligent management system and method applied to adaptive graphic rendering. BACKGROUND

[0002] The existing graphic rendering technology mainly stays in a single dimension in analyzing the device running state before starting a task, and fails to comprehensively sort out the core hardware load, resource competition degree, key rendering resource redundancy and parallel computing unit idle condition, and also lacks systematic recording and reuse of these state information. When selecting a rendering configuration, the real-time device state and historical scene data cannot be combined to predict abnormal risks, and problems such as hardware load imbalance and excessive resource competition are often caused by improper configuration, which leads to frequent device operation abnormalities, increases the failure probability, shortens the effective use cycle of the device, and affects the long-term use cost and efficiency. The traditional rendering management does not finely disassemble the whole task, and also does not divide the adaptive rendering level range according to the geometric characteristics, pixel performance, lighting demand and resource consumption of each sub-task. When determining the rendering path, it is easy to fall into an extreme choice: either blindly improve the rendering level for the pursuit of high graphic quality, leading to overburdening of the device, or excessively reduce the rendering quality to ensure the stability of the device, which cannot meet the user's demand for picture effect. This mode lacking dynamic adjustment is difficult to achieve the optimal matching of quality and efficiency within the device carrying capacity range, and the overall rendering experience is not good. The existing technology does not systematically review the whole process device state data after the completion of the rendering task, and cannot timely identify the abnormal trend of the core hardware load, resource competition and other dimensions. Even if the problem is found, it is difficult to timely prompt the device maintenance, and it is easy to cause the expansion of potential faults. At the same time, when planning subsequent tasks, the fixed historical data reference standard is still used, and the reference basis is not corrected according to the execution result of the previous task, so that the data calling is out of touch with the actual state of the current device, and it is difficult to form a closed loop of "execution-review-optimization", and it needs to rely on frequent manual intervention. In the long run, the effectiveness and reliability of the rendering management are difficult to guarantee. SUMMARY

[0003] The present application aims to provide an intelligent management system and method applied to adaptive graphic rendering to solve the problems raised in the background art.

[0004] In order to solve the above technical problems, the present application provides the following technical scheme: an intelligent management method applied to adaptive graphic rendering, comprising the following steps:

[0005] S1, analyzing the device running state to obtain a device running state label, and storing the device running state label into a historical database;

[0006] S2, disassembling the adaptive graphics rendering task to obtain a graphics rendering subtask, and dividing the graphics rendering subtask according to the workload of the graphics rendering subtask to determine the rendering intensity of the subtask rendering;

[0007] S3, considering the adaptive graphics rendering task data, predicting the device running state after the graphics rendering subtask is rendered;

[0008] S4, considering the device running state after the graphics rendering subtask is rendered, planning the rendering intensity of the graphics rendering subtask;

[0009] S5, considering the rendering intensity of the graphics rendering subtask, selecting the optimal rendering scheme;

[0010] S6, after the optimal rendering scheme is executed, modifying the effective reference period of the database according to the rendering result.

[0011] Further, in step S1, before the device performs the historical adaptive graphics rendering task, the device running state is analyzed, and a device running state label is set, the device running state label including: core hardware load saturation, hardware resource competition intensity, rendering key resource redundancy, and parallel computing unit idle rate, wherein the core hardware load saturation represents the ratio of the current hardware actual load to the rated load of the hardware, the hardware resource competition intensity represents the ratio of the resource priority of the adaptive graphics rendering task to the average resource priority of all running tasks of the device, wherein the resource priority refers to the priority level of the adaptive graphics rendering task to obtain hardware resources such as GPU and video memory, the resource priority determines the acquisition right of the adaptive graphics rendering task in resource competition, and the resource priority is obtained in the following manner: tasks are sorted from early to late according to the task receiving time point, the resource priority of the task with the earliest receiving time point is 1, and so on, the rendering key resource redundancy represents the ratio of the required resource amount of the adaptive graphics rendering task to the idle resource amount of the device, and the parallel computing unit idle rate represents the ratio of the number of unoccupied parallel computing units in the device to the total number of parallel computing units. The device running state label is stored in the historical database; through the multi-dimensional device running state label, the running characteristics of the device before starting the historical rendering task are comprehensively and accurately described, the state recognition one-sidedness caused by single-dimensional analysis is avoided, and complete device state reference is provided for subsequent rendering decision. At the same time, these label systems are stored in the historical database, effectively accumulating device running data in different scenarios, breaking the limitation of scattered and difficult to reuse of past data. When a new rendering task is received subsequently, the state data of similar historical scenarios can be quickly retrieved, and it is not necessary to analyze the device state from zero every time, which lays a solid data foundation for accurately judging the device adaptability and planning a reasonable rendering scheme, and improves the efficiency and accuracy of subsequent task decision.

[0012] Further, in step S2, before the device performs the adaptive graphics rendering task of the history, the target task is disassembled, and the target task is disassembled into M graphics rendering subtasks, wherein the graphics rendering subtask represents a rendering task of the device in a set task time period, data of the M graphics rendering subtasks is called to obtain rendering task quantities of the M graphics rendering subtasks, and the rendering task quantity is obtained in the following manner: from four dimensions of geometric data, pixel data, illumination data, and resource data: the number of vertices and triangles is counted in the geometric dimension, the resolution, the anti-aliasing level, and the number of shader instructions are combined in the pixel dimension, the number of light sources and the illumination model type are calculated in the illumination dimension, and the memory occupation is monitored in the resource dimension. Then, according to the influence weight of each dimension on rendering (for example, 30% for each of geometry and pixel), the dimension values are integrated to obtain a comprehensive quantitative value reflecting the total workload, wherein the rendering subtask workload of the mth graphics rendering subtask is A m , m = 1, 2, …, M, m = 1, 2, …, M is substituted, and the rendering subtask workloads {A m , …, A M} of the M graphics rendering subtasks are obtained, and the rendering subtask workloads corresponding to the graphics rendering subtasks are stored in the history database.

[0013] For the mth graphics rendering subtask, the geometric data, the pixel data, the illumination data, and the resource data of the rendering subtask workload are comprehensively considered to divide the subtask rendering levels of the graphics rendering subtask, and the division method of the subtask rendering level is as follows: according to the four-dimensional complexity of “geometric data-pixel data-illumination data-resource data” of the task itself, an N-level ladder system from “lowest level” to “highest level” is constructed.

[0014] The lowest level: the geometric data is a small amount of simple model, the pixel data is low resolution and has no anti-aliasing, the illumination data is only single light source + basic diffuse reflection, and the resource data demand is extremely low, such as static 2D interface.

[0015] The lower level: the geometric data contains a small amount of low model, the pixel data is medium-low resolution + basic anti-aliasing, the illumination data adds at least a large amount of light source + simple illumination data model, and the resource data demand is low, such as a lightweight 3D small game scene.

[0016] The middle level: the geometric data is a medium amount of common model, the pixel data is a conventional resolution + medium anti-aliasing, the illumination data is a medium amount of light source + a standard illumination data model, and the resource data demand is medium, such as an ordinary 3D application scene.

[0017] The higher level: the geometric data contains a large number of fine models, the pixel data is high resolution + high-order anti-aliasing, the illumination data is a large number of light sources + a physical level illumination data model, and the resource data demand is high, such as a high-quality 3D scene.

[0018] The highest level: the geometry data is extremely high-precision model, the pixel data is super high resolution + top anti-aliasing, the lighting data contains super light source + light tracing and global illumination data, and the resource data requires extremely high resources, such as the complex scene of a top 3A game.

[0019] N can be adjusted as needed, and can be achieved by splitting or merging adjacent levels to maintain the step-by-step progression of the complexity description of each dimension, and the upper limit of the sub-task rendering level of the mth graphics rendering sub-task is B m , wherein the geometry includes the number of rendering vertices and the number of triangles, the pixel includes the clarity and the anti-aliasing capability, the lighting includes the number of light sources and the model category, and the resource includes the amount of video memory, and substituting m = 1, 2, …, M, the upper limit of the sub-task rendering level of the M graphics rendering sub-tasks is {B1, B2, …, B m , …, B M}, and the upper limit of the level B m indicates that the rendering level is provided between 1 and B m when the mth graphics rendering sub-task is performed; by splitting the overall rendering task into sub-tasks and matching the workload of the geometry data, the pixel data, the lighting data and the resource data in multiple dimensions, the general judgment of the task load is avoided, and the actual work demand of each sub-task is accurately presented, thereby providing a clear basis for subsequent rendering level planning. Meanwhile, the step-by-step rendering level constructed according to the four-dimensional complexity of the sub-task, combined with the clear upper limit of the level, allows the rendering configuration of each sub-task to have an adaptive selection range, thereby avoiding the blindness of the configuration and flexibly matching the complexity demand of different sub-tasks. In addition, the sub-task information, the workload and the level data are stored in the database, historical references are effectively accumulated, repeated work for subsequent analysis and planning of new tasks is saved, and the overall management efficiency is improved.

[0020] Further, in step S3, when the adaptive graphics rendering task is received, the adaptive graphics rendering task is analyzed, the rendering sub-task workload of the M graphics rendering sub-tasks is obtained {A1, A2, …, A m , …, A M}, and then the rendering sub-task level planning is performed, the device running state label at the time when the adaptive graphics rendering task is received is obtained as a real-time label, the mth graphics rendering sub-task is planned, the real-time label before the device processes the mth graphics rendering sub-task is obtained, a preset reference time period T is selected with the current time point as the end point, the data of the sub-task processing at the same time when the device running state label is the same as the real-time label is obtained, wherein the number of times of the sub-task processing with the rendering level C is Q m , C = 1, 2, …, B m , the data of the Q m sub-task processing is analyzed, and Q m is called.The running state of the device after processing the second sub-task, Q m The device running state label with the highest occurrence frequency in the running states is taken as the real-time label before the (m+1)th graphic rendering sub-task. The core hardware load saturation, hardware resource competition intensity, rendering key resource redundancy, and parallel computing unit idle rate are set as the device normal running range. If any of the core hardware load saturation, hardware resource competition intensity, rendering key resource redundancy, and parallel computing unit idle rate is not in the device normal running range, it is determined that the device is running abnormally. Otherwise, it is determined that the device is running normally, and the number of times of device running abnormally after processing the second sub-task Q m m is obtained, and the expected device running abnormality K m after rendering of the (m)th graphic rendering sub-task at rendering level C is obtained. m m m C=1,2,…,B m , the expected device running abnormality after rendering of the (m)th graphic rendering sub-task at B m rendering levels is obtained. The device abnormality risk is accurately predicted. By matching the real-time device state label of the new task with the data of the same scene in the historical database and combining the explicit device normal running range, the system analyzes the device state after processing the sub-task at different rendering levels, obtains the abnormality expectation corresponding to each level, and avoids the device sudden abnormality caused by the experience-based selection of levels.

[0021] Further, in step S4, m=1,2,…,M is substituted, the device running abnormality expectation from the 1st graphic rendering sub-task to the Mth rendering sub-task is obtained, Y adaptive graphic rendering task rendering paths are acquired, the rendering path of the adaptive graphic rendering task is determined by the rendering level selected by the sub-task, and the device running abnormality prediction value D y of the yth adaptive graphic rendering task rendering path is obtained, y=1,2,…,Y. The device running abnormality prediction value of the yth adaptive graphic rendering task rendering path represents the sum of the device running abnormality expectations from the 1st graphic rendering sub-task to the Mth rendering sub-task in the yth adaptive graphic rendering task rendering path. y=1,2,…,Y is substituted, the device running abnormality prediction values {D1,D2,…,D y ,…,D Y} of the Y adaptive graphic rendering task rendering paths are obtained. A device running abnormality rate threshold Z is set. If D y ​​​ / M) > Z, it is judged that the rendering path of the yth adaptive graphics rendering task is abnormal; otherwise, it is judged that the rendering path of the yth adaptive graphics rendering task is normal, the rendering path of the yth adaptive graphics rendering task is taken as a candidate path, and then X candidate paths are obtained; by integrating the device operation abnormality expectations of each subtask, the risk assessment of the whole rendering path is formed, the limitation of single subtask analysis is broken through, the device operation risk of the whole path can be comprehensively judged, the decision deviation caused by local rationality but overall abnormality is avoided, and the interference of invalid path on decision is reduced.

[0022] Further, in step S5, X candidate paths are analyzed, and the optimal indexes of the X candidate paths are {E1, E2, …, E x ,…,E X}, wherein E x represents the optimal index of the xth candidate path, the optimal index of the xth candidate path is the sum of the grades of the M graphics rendering subtasks of the xth candidate path, and then the path with the highest optimal index is selected to render the target task.

[0023] Further, in step S6, after the target task is rendered, the device operation state after any rendering subtask is completed is searched, the number K1 of device operation states in which the core hardware load saturation degree is not in the normal operation range of the device, the number K2 of device operation states in which the hardware resource competition intensity is not in the normal operation range of the device, the number K3 of device operation states in which the rendering key resource redundancy degree is not in the normal operation range of the device, and the number K4 of device operation states in which the parallel computing unit idle rate is not in the normal operation range of the device are obtained, if K1 / M > Z, it is judged that the core hardware load saturation degree is normal; otherwise, it is judged that the core hardware load is abnormal, and then the hardware resource competition intensity, the rendering key resource redundancy degree, and the parallel computing unit idle rate are judged one by one, when any device operation state label is abnormal, it is prompted to maintain the device, and the reference time period is corrected, the reference time period is corrected to F*T, F = 1-(K max / M) + Z, K max is the maximum value in K1, K2, K3, and K4, when the next adaptive graphics rendering task is received, the database is screened by using the corrected reference time period, until the device maintenance is monitored to be completed, the reference time period is reset to T; the device state is reviewed in time after the rendering task is completed, potential problems in hardware load, resource competition, etc. are accurately identified by checking the abnormality of each dimension operation index, once the abnormality is found, the maintenance is prompted in time, the small problem is avoided from being expanded into a serious fault, the reference period of historical data is dynamically corrected according to the abnormality, the subsequent task screening data is more in line with the actual state of the current device, and the interference of outdated data on decision is reduced.

[0024] The application relates to an intelligent management system applied to adaptive graphics rendering, and the system comprises a device state analysis module, a subtask intensity division module, a rendering state prediction module, a rendering intensity planning module, an optimal scheme selection module and a reference cycle correction module.

[0025] The device state analysis module is used for analyzing a device running state to obtain a device running state label, and the device running state label is stored in a historical database.

[0026] The subtask intensity division module is used for disassembling an adaptive graphics rendering task to obtain a graphics rendering subtask, and the rendering intensity of subtask rendering is divided according to the workload of the graphics rendering subtask.

[0027] The rendering state prediction module is used for predicting a device running state after a graphics rendering subtask is rendered by considering adaptive graphics rendering task data.

[0028] The rendering intensity planning module is used for planning the rendering intensity of a graphics rendering subtask by considering the device running state after the graphics rendering subtask is rendered.

[0029] The optimal scheme selection module is used for selecting an optimal rendering scheme by considering the rendering intensity of a graphics rendering subtask.

[0030] The reference cycle correction module is used for modifying the effective reference cycle of a database according to a rendering result after an optimal rendering scheme is executed.

[0031] Compared with the prior art, the application has the beneficial effects that: on the one hand, the multi-dimensional analysis of the device running state before the rendering task is started is recorded to form historical data for reference. When a new rendering task is received, the historical data of the past processing under the matching scene is screened from the historical data in combination with the real-time device state to accurately judge the possibility of device abnormality under different rendering configurations. By identifying the rendering selection that may cause the device state to exceed the normal range in advance, the improper configuration is actively avoided, the device abnormality caused by problems such as hardware load imbalance and excessive resource competition is effectively avoided, the hardware is ensured to be always in a stable running interval, the fault risk is reduced, and the effective use cycle of the device is prolonged; on the one hand, the overall rendering task is divided into multiple sub-tasks, a reasonable rendering level range is divided for each sub-task, and the rendering scheme that can be borne by the device is screened based on the historical data. When the final rendering path is determined, the path with a higher rendering level is preferentially selected under the premise of ensuring the normal operation of the device, which avoids excessively reducing the rendering quality for the purpose of device stability and prevents the device from being overburdened due to blind improvement of the level. This dynamic adjustment mode makes the rendering effect optimal within the bearing capacity of the device, balances the demand of users for the quality of graphics and the running efficiency of the device, and improves the overall rendering experience; on the other hand, after the rendering task is completed, the method reviews the device state data in the execution process to determine whether the running state of each dimension is normal. If it is found that a certain type of state is frequently abnormal, the device maintenance is prompted in time to avoid the expansion of potential faults. At the same time, the reference period of the historical data screened in the subsequent review is dynamically corrected according to the review result, so that the historical data calling is more in line with the actual state of the device, and the accuracy of the subsequent rendering planning is improved. This virtuous cycle of “execution-review-optimization” enables the method to continuously adapt to the changes of the device state, and long-term guarantees the effectiveness and reliability of the rendering management. BRIEF DESCRIPTION OF DRAWINGS

[0032] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application together with the embodiments thereof, and are used to explain the application, and do not constitute a limitation on the application. In the drawings:

[0033] Figure 1 is a structural diagram of an intelligent management system applied to adaptive graphics rendering of the application;

[0034] Figure 2 is a flowchart of an intelligent management method applied to adaptive graphics rendering of the application. DETAILED DESCRIPTION

[0035] Clearly, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0036] Please refer to Figure 1 and Figure 2 The present application provides a technical solution: an intelligent management method applied to adaptive graphics rendering, comprising the following steps:

[0037] S1, analyzing the device running state to obtain a device running state label, and storing the device running state label in a historical database;

[0038] S2, disassembling an adaptive graphics rendering task to obtain a graphics rendering subtask, and dividing the rendering intensity of the subtask rendering by considering the workload of the graphics rendering subtask;

[0039] S3, considering adaptive graphics rendering task data, predicting the device running state after rendering the graphics rendering subtask;

[0040] S4, considering the device running state after rendering the graphics rendering subtask, planning the rendering intensity of rendering the graphics rendering subtask;

[0041] S5, considering the rendering intensity of rendering the graphics rendering subtask, selecting an optimal rendering scheme;

[0042] S6, after the optimal rendering scheme is executed, modifying the effective reference period of the database according to the rendering result.

[0043] In step S1, before the device performs the adaptive graphics rendering task in the history, the device running state is analyzed, and a device running state label is set, the device running state label including: core hardware load saturation, hardware resource competition intensity, rendering key resource redundancy, and parallel computing unit idle rate, wherein the core hardware load saturation represents a ratio of a current hardware actual load to a hardware rated load, the hardware resource competition intensity represents a ratio of a resource priority of the adaptive graphics rendering task to an average resource priority of all current running tasks of the device, the rendering key resource redundancy represents a ratio of a required resource amount of the adaptive graphics rendering task to a free resource amount of the device, and the parallel computing unit idle rate represents a ratio of a number of unoccupied parallel computing units in the device to a total number of parallel computing units, and the device running state label is stored in a history database; through the multi-dimensional device running state label, the running characteristics of the device before starting the history rendering task are comprehensively and accurately described, the state recognition one-sidedness caused by single-dimensional analysis is avoided, and complete device state references are provided for subsequent rendering decisions. Meanwhile, these label systems are stored in the history database, device running data in different scenes are effectively accumulated, and the limitation that past data is scattered and difficult to reuse is broken. When a new rendering task is received subsequently, the state data of similar historical scenes can be quickly retrieved, and it is not necessary to analyze the device state from zero every time, so that a solid data foundation is laid for accurately judging device adaptability and planning a reasonable rendering scheme, and the efficiency and accuracy of subsequent task decisions are improved.

[0044] In step S2, before the device performs the adaptive graphics rendering task in the history, the target task is disassembled, and the target task is disassembled into M graphics rendering subtasks, wherein the graphics rendering subtask represents a rendering task of the device in a set task time period, data of the M graphics rendering subtasks is called, and rendering task amounts of the M graphics rendering subtasks are obtained, the rendering task amount is obtained in the following four dimensions: geometry, pixel, light, and resource, the geometry dimension is used to count the number of vertices / triangles, the pixel dimension is used to count the resolution, the anti-aliasing level, and the number of shader instructions, the light dimension is used to count the number of light sources and the light model type, and the resource dimension is used to monitor the video memory occupation. Then, according to the influence weight of each dimension on rendering (for example, 30% for geometry and 30% for pixel), the values of each dimension are integrated to obtain a comprehensive quantitative value reflecting the total workload, wherein the rendering subtask workload of the mth graphics rendering subtask is A m , m = 1, 2, …, M, m = 1, 2, …, M is substituted to obtain the rendering subtask workloads {A1, A2, …, A m , …, A M} of the M graphics rendering subtasks, and the rendering subtask workloads are stored in the history database corresponding to the graphics rendering subtasks.

[0045] For the mth graphics rendering subtask, the subtask rendering level of the graphics rendering subtask is divided by comprehensively considering the geometric data, pixel data, illumination data and resource data of the rendering subtask workload. The subtask rendering level division method is: according to the four-dimensional complexity of "geometric data-pixel data-illumination data-resource data" of the task itself, an N-level ladder system from "lowest level" to "highest level" is constructed:

[0046] The lowest level: the geometric data is a small amount of simple model, the pixel data is low resolution and no anti-aliasing, the illumination data is only single light source + basic diffuse reflection, and the resource data demand is extremely low, such as static 2D interface.

[0047] Lower level: the geometric data contains a small amount of low model, the pixel data is medium-low resolution + basic anti-aliasing, the illumination data adds at least light source + simple illumination data model, and the resource data demand is low, such as lightweight 3D small game scene.

[0048] Middle level: the geometric data is a medium number of common models, the pixel data is a regular resolution + medium anti-aliasing, the illumination data is a medium number of light sources + a standard illumination data model, and the resource data demand is medium, such as a common 3D application scene.

[0049] Higher level: the geometric data contains a large number of fine models, the pixel data is high resolution + high-order anti-aliasing, the illumination data is more light sources + physical level illumination data model, and the resource data demand is high, such as high-quality 3D scene.

[0050] The highest level: the geometric data is a large number of high-precision models, the pixel data is super high resolution + top anti-aliasing, the illumination data contains super light source + light tracing and global illumination data, and the resource data demand is extremely high, such as a top 3A game complex scene.

[0051] N can be adjusted as needed, and adjacent levels can be split or combined to maintain the ladder progression of the complexity description of each dimension. The upper limit of the subtask rendering level of the mth graphics rendering subtask is B m , where geometry includes the number of rendering vertices and triangles, pixels include clarity and anti-aliasing capability, illumination includes the number of light sources and model categories, and resources include the level of video memory. Substitute m = 1, 2, …, M to get the upper limit of the subtask rendering level of the M graphics rendering subtasks {B1, B2, …, BM} m , and the upper limit of the level B M . m represents providing choices 1 to B mThe rendering levels are defined by breaking down the overall rendering task into subtasks and quantifying workload across multiple dimensions, including geometric, pixel, lighting, and resource data. This avoids vague assessments of task load and accurately presents the actual workload requirements of each subtask, providing a clear basis for subsequent rendering level planning. Furthermore, the tiered rendering levels, constructed based on the four dimensions of subtask complexity and with clear level caps, allow for a suitable range of rendering configurations for each subtask, avoiding blind configuration and flexibly matching the complexity requirements of different subtasks. In addition, storing subtask information, workload, and level data in a database effectively accumulates historical references, saving repetitive work for the analysis and planning of new tasks and improving overall management efficiency.

[0052] In step S3, upon receiving an adaptive graphics rendering task, the adaptive graphics rendering task is analyzed to obtain the rendering subtask workloads {A1, A2, ..., A} of the M graphics rendering subtasks. m ,…,A M The process then proceeds to plan the rendering subtask levels, obtaining the device operating status label when the adaptive graphics rendering task is received as the real-time label. For the m-th graphics rendering subtask, subtask planning is performed, obtaining the real-time label before the device processes the m-th graphics rendering subtask. A preset reference time period T, with the current time as the endpoint, is used to filter the database, obtaining the subtask processing data at the same time when the device operating status label and the real-time label are the same. The number of times the subtask with rendering level C is Q is recorded. m C=1,2,…,B m , for Q m Analyze the data processed by the secondary task and call Q. m The operating status of the device after the secondary task is processed, Q m The device running status label that appears most frequently in each running state is used as the real-time label before the (m+1)th graphics rendering subtask. Normal operating ranges are set for core hardware load saturation, hardware resource contention intensity, rendering critical resource redundancy, and parallel computing unit idle rate. If any of these values ​​is outside the normal operating range, the device is considered abnormal; otherwise, it is considered normal, and the call to Q is obtained. m The number of times the device malfunctions after the secondary task is processed is q. m This leads to the expected device operation anomaly K after rendering the m-th graphics rendering subtask at rendering level C. m K m =q m / Q m Substitute them one by one into C=1,2,…,B m, obtain the device operation abnormality expectation of the mth graphics rendering subtask rendered at B m level, accurately predict the device abnormality risk, match the real-time device state label of the new task with the data of the same scene in the historical database, combine the explicit device normal operation range, analyze the device state after the subtask processing at different rendering levels, obtain the abnormality expectation corresponding to each level, and avoid the device sudden abnormality caused by the experience-based selection of the level.

[0053] In step S4, m = 1, 2, …, M is substituted one by one to obtain the device operation abnormality expectation from the first graphics rendering subtask to the Mth rendering subtask, obtain Y adaptive graphics rendering task rendering paths, and the rendering path of the adaptive graphics rendering task is determined by the rendering level selected by the subtask, and then obtain the device operation abnormality prediction value D y of the yth adaptive graphics rendering task rendering path, y = 1, 2, …, Y, the device operation abnormality prediction value of the yth adaptive graphics rendering task rendering path represents the sum of the device operation abnormality expectations from the first graphics rendering subtask to the Mth rendering subtask in the yth adaptive graphics rendering task rendering path, and m = 1, 2, …, M is substituted one by one to obtain the device operation abnormality prediction value {D1, D2, …, D y , …, D Y} of the Y adaptive graphics rendering task rendering paths, set the device operation abnormality rate threshold Z, if D y / M > Z, it is judged that the yth adaptive graphics rendering task rendering path is abnormal, otherwise it is judged that the yth adaptive graphics rendering task rendering path is normal, the yth adaptive graphics rendering task rendering path is taken as a candidate path, and then X candidate paths are obtained; by integrating the device operation abnormality expectations of each subtask, the risk assessment of the whole rendering path is formed, which breaks through the limitation of single subtask analysis, can comprehensively judge the device operation risk of the whole path, avoids the decision deviation caused by the local rationality but the whole abnormality, and reduces the interference of invalid paths on the decision.

[0054] In step S5, the X candidate paths are analyzed, and the optimal index of the X candidate paths is {E1, E2, …, E x , …, E X}, wherein E x represents the optimal index of the xth candidate path, the optimal index of the xth candidate path is the sum of the levels of the M graphics rendering subtasks of the xth candidate path, and then the path with the highest optimal index is selected to render the target task.

[0055] In step S6, after rendering the target task is completed, the device running state after any rendering subtask is completed is retrieved, the number K1 of device running states when the core hardware load saturation is not in the normal running range of the device, the number K2 of device running states when the hardware resource competition intensity is not in the normal running range of the device, the number K3 of device running states when the rendering key resource redundancy is not in the normal running range of the device, and the number K4 of device running states when the parallel computing unit idle rate is not in the normal running range of the device are obtained. If K1 / M>Z, it is judged that the core hardware load saturation is normal. Otherwise, it is judged that the core hardware load is abnormal, and then the hardware resource competition intensity, the rendering key resource redundancy, and the parallel computing unit idle rate are judged one by one. When any device running state label is abnormal, it is prompted to maintain the device, and the reference time period is corrected. The reference time period is corrected as F*T, F=1-(K max / M)+Z, K max is the maximum value in K1, K2, K3, and K4. When the next adaptive graphics rendering task is received, the database is screened by using the corrected reference time period until the device maintenance is completed. The reference time period is reset as T. After the rendering task is completed, the device state is reviewed in time. The potential problems in the hardware load and resource competition are accurately identified by checking the abnormal conditions of each dimension running index. Once the abnormality is found, it is promptly maintained to avoid the small problem from expanding into a serious failure. The reference period of the historical data is dynamically corrected according to the abnormal condition, so that the subsequent task screening data is more in line with the actual state of the current device, and the interference of outdated data on the decision is reduced.

[0056] An intelligent management system applied to adaptive graphics rendering, the system comprising: a device state analysis module, a subtask intensity division module, a rendering state prediction module, a rendering intensity planning module, an optimal scheme selection module, and a reference period correction module;

[0057] The device state analysis module is used for analyzing the device running state to obtain a device running state label, and storing the device running state label into a historical database;

[0058] The subtask intensity division module is used for decomposing an adaptive graphics rendering task to obtain a graphics rendering subtask, and dividing the rendering intensity of the subtask rendering by considering the workload of the graphics rendering subtask;

[0059] The rendering state prediction module is used for predicting the device running state after rendering the graphics rendering subtask by considering the adaptive graphics rendering task data;

[0060] The rendering intensity planning module is used for planning the rendering intensity of the graphics rendering subtask by considering the device running state after rendering the graphics rendering subtask;

[0061] The optimal scheme selection module is configured to select an optimal rendering scheme in consideration of rendering intensity of the graphic rendering subtask.

[0062] The reference period correction module is configured to correct the effective reference period of the database according to the rendering result after the optimal rendering scheme is executed.

[0063] In step S1, before starting the historical rendering task of the 3D game, the running state of the device is comprehensively analyzed. The actual load of the core hardware such as the graphics card and the CPU is checked through the hardware monitoring tool to determine the core hardware load saturation. The resource priority of the current device running background program, other applications and game tasks is counted to compare the hardware resource competition intensity. The proportion of key resources such as video memory and computing power required for game rendering and idle resources of the device is calculated to obtain the rendering key resource redundancy. The number of GPU parallel computing units not occupied in the device is investigated, and the proportion of the total parallel computing units is calculated to determine the parallel computing unit idle rate. The four indexes are integrated into the device running state label and stored in the historical database in real time to retain the scene state data for subsequent tasks.

[0064] After entering step S2, the overall rendering task of the 3D game is first decomposed into four subtasks of scene modeling rendering, character animation rendering, light and shadow effect rendering, and particle special effect rendering, each corresponding to a fixed task time period. For each subtask, the rendering workload is quantified from multiple dimensions: the scene modeling rendering subtask counts the number of vertices and triangles of the model in the scene (geometric dimension), combines the resolution, antialiasing level and number of shader instructions set by the game (pixel dimension), calculates the number of light sources in the scene and the type of lighting model used (lighting dimension), monitors the video memory occupation required by the subtask (resource dimension), and then integrates the comprehensive workload of each subtask according to the influence weight of each dimension (such as higher geometric and pixel dimension weight). Then, the subtask rendering level is divided according to the four-dimensional complexity, for example, if the model is simple, the resolution is low, and there is only a single light source, the scene modeling rendering subtask is divided into the lowest level; if the model is fine, the resolution is high, there are multiple light sources, and the physical level lighting model is used, the scene modeling rendering subtask is divided into a higher level, and the rendering level upper limit of each subtask is determined. The subtask information, workload and level data are stored in the database.

[0065] When the user starts the 3D game rendering task again, step S3 is executed to analyze the subtask composition and workload of the current task and match them with the historical subtask data in the database. At the same time, real-time running state labels of the device are obtained, such as the current graphics card load, background resource occupation, etc. The historical data consistent with the real-time labels is filtered from the database with the current time as the end point and according to the preset reference period. Taking the character animation rendering subtask as an example, the processing records of different rendering levels of the subtask in the historical data are extracted, and it is checked whether the device state after each processing is within the normal range (such as the core hardware load saturation not exceeding 80%, and the resource competition intensity being below moderate). The proportion of the number of device exceptions to the total number of processing times at each level is calculated to obtain the device running exception expectation of different rendering levels of the subtask. The same is true for other subtasks.

[0066] In step S4, the exception expectations of all subtasks are integrated to generate multiple rendering paths, each path being composed of the rendering levels of each subtask. For example, one path is “scene modeling low level + character animation medium level + light and shadow effect medium level + particle special effect low level”, and another path is “scene modeling medium level + character animation high level + light and shadow effect high level + particle special effect medium level”. The device running exception prediction value (i.e. the sum of the exception expectations of each subtask) of each path is calculated, a device running exception rate threshold (such as an exception rate not exceeding 1%) is set, and the selected paths whose exception prediction values meet the threshold requirements are screened out to exclude paths that may cause frequent device exceptions.

[0067] In step S5, the preferred index of each selected path is calculated, the rendering levels of all subtasks in the path are added, and the path with the highest preferred index means that the overall rendering quality is better. For example, if the sum of the rendering levels of each subtask in a selected path is higher than that of other paths, and its exception rate meets the requirements, then the path is selected as the final rendering scheme, and the game rendering is started.

[0068] After the rendering task is completed, step S6 is entered to retrieve the device state data after each subtask is completed, and the number of times that the core hardware load saturation and the hardware resource competition intensity exceed the normal range are counted. If it is found that the number of times of core hardware load saturation exception is large, the user is prompted in time to perform device maintenance (such as cleaning the dust of the graphics card, and closing the background high-occupancy program); at the same time, according to the number of times of the most frequent exception dimension, the reference period for filtering historical data is corrected. If the number of times is large, the reference period is shortened, so that the historical data called later is more consistent with the current device state. After the user completes the device maintenance and monitors that the device state returns to normal, the reference period is reset to the initial value.

[0069] It will be apparent to those skilled in the art that the application is not limited to the details of the above-exemplified sensor device embodiments, but that the application can be implemented in other concrete forms without departing from the spirit or essential characteristics thereof. The embodiments should, therefore, be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. No reference signs in the claims should be considered as limiting the scope of the claims with respect to the figures of the patent document.

Claims

1. A method for intelligent management applied to adaptive graphics rendering, characterized in that: The method comprises the following steps: S1, obtaining a device running state label by analyzing a device running state and storing the device running state label in a historical database; S2, obtaining a graphic rendering subtask by disassembling an adaptive graphic rendering task and dividing rendering intensity of subtask rendering according to workload of the graphic rendering subtask; S3, predicting a device running state after rendering of the graphic rendering subtask according to adaptive graphic rendering task data; S4, planning rendering intensity of the graphic rendering subtask according to the device running state after rendering of the graphic rendering subtask; S5, selecting an optimal rendering scheme according to the rendering intensity of the graphic rendering subtask; S6, modifying an effective reference period of a database according to a rendering result after execution of the optimal rendering scheme is completed. In step S3, when receiving the adaptive graphics rendering task, the adaptive graphics rendering task is analyzed to obtain rendering subtask workloads {A1, A2, …, A m ,…,A M} of M graphics rendering subtasks, and then rendering subtask level planning is performed on the rendering subtasks, a device running state label at the time when the adaptive graphics rendering task is received is obtained as a real-time label, the mth graphics rendering subtask is planned, a real-time label before the device processes the mth graphics rendering subtask is obtained, a preset reference time period T with the current time point as an endpoint is screened, data of subtask processing at the same time when the device running state label is the same as the real-time label is obtained, wherein the number of times of subtask processing of the rendering level C is Q m , C=1,2,…,B m , the data of Q m times of subtask processing is analyzed, the running state of the device after Q m times of subtask processing is called, the device running state label that appears most frequently in Q m running states is taken as the real-time label before the m+1th graphics rendering subtask, the core hardware load saturation, the hardware resource competition intensity, the rendering key resource redundancy, and the parallel computing unit idle rate are respectively set as the device normal running range, when any one of the core hardware load saturation, the hardware resource competition intensity, the rendering key resource redundancy, and the parallel computing unit idle rate is not in the device normal running range, it is judged that the device is running abnormally; otherwise, it is judged that the device is running normally, and then the number of times of device running abnormally after Q m times of subtask processing is obtained q m , and then the device running abnormal expectation K m after the rendering of the mth graphics rendering subtask at the rendering level C is obtained, K m =q m / Q m , C=1,2,…,B m is substituted, and the device running abnormal expectation after the rendering of the mth graphics rendering subtask at B m rendering levels is obtained. 2.The intelligent management method applied to adaptive graphics rendering according to claim 1, characterized in that: In step S1, before a historical adaptive graphic rendering task is performed by a device, a device running state is analyzed, and a device running state label is set, wherein the device running state label comprises core hardware load saturation, hardware resource competition intensity, rendering key resource redundancy, and parallel computing unit idle rate, wherein the core hardware load saturation represents a ratio of a current hardware actual load to a hardware rated load, the hardware resource competition intensity represents a ratio of a resource priority of the adaptive graphic rendering task to an average resource priority of all running tasks of the device, the rendering key resource redundancy represents a ratio of a required resource amount of the adaptive graphic rendering task to an idle resource amount of the device, and the parallel computing unit idle rate represents a ratio of a number of unoccupied parallel computing units in the device to a total number of parallel computing units, and the device running state label is stored in the historical database.

3. The intelligent management method applied to adaptive graphics rendering according to claim 2, characterized in that: In step S2, before the device performs the adaptive graphics rendering task of the history, the target task is disassembled, and the target task is disassembled into M graphics rendering subtasks, wherein the graphics rendering subtask represents the rendering task of the device in a certain task time period, data of the M graphics rendering subtasks is called to obtain rendering task amounts of the M graphics rendering subtasks, wherein the rendering subtask workload of the mth graphics rendering subtask is A m , m = 1, 2, …, M, m = 1, 2, …, M is substituted to obtain rendering subtask workloads {A1, A2, …, A m , …, A M} of the M graphics rendering subtasks, and the rendering subtask workloads are stored in the history database corresponding to the graphics rendering subtasks.

4. The intelligent management method applied to adaptive graphics rendering according to claim 3, characterized in that: For the mth graphics rendering subtask, considering the geometry data, pixel data, illumination data and resource data of the rendering subtask workload, the subtask rendering level of the graphics rendering subtask is divided, and the upper limit of the subtask rendering level of the mth graphics rendering subtask is B m , wherein the geometry data includes the number of rendering vertices and the number of triangles, the pixel data includes the clarity and the anti-aliasing capability, the illumination data includes the number of light sources and the model category, and the resource data includes the amount of video memory; and substituting m = 1, 2, …, M, the upper limits of the subtask rendering levels of the M graphics rendering subtasks are obtained as {B1, B2, …, BM} m . M , and the upper limit of the level B m represents that the rendering level between 1 and B m is provided when the mth graphics rendering subtask is performed.

5. The intelligent management method applied to adaptive graphics rendering according to claim 4, characterized in that: In step S4, m=1, 2, …, M is substituted one by one to obtain the device operation abnormality expectation from the first graphics rendering subtask to the Mth rendering subtask, obtain Y adaptive graphics rendering task rendering paths, the adaptive graphics rendering task rendering path is determined by the rendering level selected by the subtask processing, and then obtain the device operation abnormality prediction value D of the yth adaptive graphics rendering task rendering path y , y=1, 2, …, Y, the device operation abnormality prediction value of the yth adaptive graphics rendering task rendering path represents the sum of the device operation abnormality expectation from the first graphics rendering subtask to the Mth rendering subtask in the yth adaptive graphics rendering task rendering path, m=1, 2, …, M is substituted one by one to obtain the device operation abnormality prediction value {D1, D2, …, D y ,…,D Y} of the Y adaptive graphics rendering task rendering paths, set a device operation abnormality rate threshold Z, if D y / M>Z, it is judged that the yth adaptive graphics rendering task rendering path is abnormal; otherwise, it is judged that the yth adaptive graphics rendering task rendering path is normal, the yth adaptive graphics rendering task rendering path is taken as a to-be-selected path, and then X to-be-selected paths are obtained.

6. The intelligent management method applied to adaptive graphics rendering according to claim 5, characterized in that: In step S5, X candidate paths are analyzed, and the preference indexes of the X candidate paths are {E1, E2, …, Ex}, wherein Ex represents the preference index of the xth candidate path, the preference index of the xth candidate path is the sum of the ranks of the M graphic rendering sub-tasks of the xth candidate path, and then the path with the highest preference index is selected to render the target task. x X x , …, Ex}, wherein Ex represents the preference index of the xth candidate path, the preference index of the xth candidate path is the sum of the ranks of the M graphic rendering sub-tasks of the xth candidate path, and then the path with the highest preference index is selected to render the target task.​​ 7. The intelligent management method applied to adaptive graphics rendering according to claim 6, characterized in that: In step S6, after rendering the target task is completed, retrieve any rendering sub task after the completion of the device running state, get the core hardware load saturation is not in the normal operation of the device when the number of device running state K1, hardware resource competition intensity is not in the normal operation of the device when the number of device running state K2, rendering key resource redundancy is not in the normal operation of the device when the number of device running state K3 and parallel computing unit idle rate is not in the normal operation of the device when the number of device running state K4, if K1 / M>Z, the core hardware load saturation is normal; otherwise, the core hardware load is abnormal, and then the hardware resource competition intensity, rendering key resource redundancy and parallel computing unit idle rate are judged one by one. When any device running state label is abnormal, prompt to maintain the device, and correct the reference time period. The reference time period is corrected as F*T, F=1-(K max / M)+Z, K max is the maximum value of K1, K2, K3 and K4. When receiving the next adaptive graphics rendering task, the database is screened by using the corrected reference time period, until the device maintenance is completed, and the reference time period is reset to T.

8. An intelligent management system applied to adaptive graphics rendering, the system being applied to the intelligent management method applied to adaptive graphics rendering according to any one of claims 1-7, characterized in that: The system comprises a device state analysis module, a subtask intensity division module, a rendering state prediction module, a rendering intensity planning module, an optimal scheme selection module, and a reference period correction module. The device state analysis module is configured to analyze a device running state to obtain a device running state label and store the device running state label in a historical database. The subtask intensity division module is configured to obtain a graphic rendering subtask by disassembling an adaptive graphic rendering task and divide rendering intensity of subtask rendering according to workload of the graphic rendering subtask. The rendering state prediction module is configured to predict a device running state after rendering of the graphic rendering subtask according to adaptive graphic rendering task data. The rendering intensity planning module is configured to plan rendering intensity of the graphic rendering subtask according to the device running state after rendering of the graphic rendering subtask. The optimal scheme selection module is configured to select an optimal rendering scheme according to the rendering intensity of the graphic rendering subtask. The reference period correction module is configured to modify an effective reference period of a database according to a rendering result after execution of the optimal rendering scheme is completed.

Images