A Scheduling Optimization Method for Parallel Processing of Head Data and GPU

Abstract

This application relates to the technical field of data scheduling, and specifically relates to a scheduling optimization method for parallel processing of nose data and GPU. The method includes: obtaining all the data in the GPU global memory and the data to be analyzed in the nose commands; updating the priority level of the hash key values in the cache data table of the GPU shared memory according to the distribution of the hash key values of all the data and the real-time access situation in the cache data table; and updating the cache data table in the GPU shared memory by using the priority level of the hash key values of the data to be analyzed. This application optimizes the scheduling of parallel processing of nose data and GPU, avoids the problem of low data scheduling efficiency in the prior art, and can effectively improve the data processing efficiency and GPU operation efficiency.

Description

Technical Field

[0001] This application relates to the technical field of data scheduling, and specifically relates to a scheduling optimization method for parallel processing of head data and GPU. Background Art

[0002] Head data, as the core input at the front end of the server, carries a rich variety of requests initiated by users, and these data are input into the GPU of the server. The GPU divides the data into countless small pieces through parallel processing technology and processes them simultaneously. This parallel processing ability not only greatly improves the speed and efficiency of data processing, but also maintains the stability and response speed of the server in the face of a large number of concurrent requests, ensuring the smooth flow of the data stream. In addition, GPU parallel processing helps to achieve load balancing by intelligently allocating tasks to different GPU cores, improving the processing ability and reliability of the entire system. When a core fails, other cores can quickly take over the tasks to ensure the continuity of data processing and the stability of the system.

[0003] For the parallel processing of head data and GPU, during the processing, some data will be stored in the shared memory for data caching, so that the GPU can quickly schedule during the multi-threaded parallel processing. The traditional technology stores the shared memory by forming a FIFO queue of data, and each update is the head of the linked list, and the tail of the linked list is eliminated. This method is likely to cause data with a relatively low frequency to be stored in the cache, occupying the cache space. When the GPU multi-threads schedule data, it needs to traverse this data, affecting the time for multi-threads to schedule data, and ultimately resulting in low data processing efficiency and GPU operation efficiency. Summary of the Invention

[0004] To solve the above technical problems, this application provides a scheduling optimization method for parallel processing of head data and GPU to solve the existing problems.

[0005] A scheduling optimization method for parallel processing of head data and GPU in this application adopts the following technical solutions:

[0006] An embodiment of this application provides a scheduling optimization method for parallel processing of head data and GPU, and this method includes the following steps:

[0007] Step 1: Obtain all the data in the GPU global memory and the data to be analyzed in the head command;

[0008] Step 2: Update the priority level of the hash key values in the cache data table of the GPU shared memory according to the distribution of the hash key values of all the data and the real-time access situation in the cache data table; specifically:

[0009] A1. Calculate the mapping relationship conflict amount of each hash key value by analyzing the number of data types and the data chaos that occur under the same hash key value for all data.

[0010] A2. Combine the mapping relationship conflict amount, the probability of the data with the highest frequency of occurrence among all data for each hash key value, and the maximum value of the frequency of occurrence of each hash key value among all its data types to calculate the arrangement priority level of each hash key value; screen out a preset number of hash key values according to the size of the arrangement priority level and store them in the cache data table in the GPU shared memory.

[0011] A3. Analyze the average level of the difference values after differentiating all hash key values in the real-time cache data table according to the time sequence when they enter the cache data table, and the average level of the difference values after differentiating according to the arrangement priority level of all hash key values, and calculate the priority level update amount of the hash key values in the cache data table.

[0012] A4. Use the priority level update amount to update the arrangement priority level of the hash key values in the cache data table.

[0013] Step 3: Update the cache data table in the GPU shared memory using the arrangement priority level of the hash key values of the data to be analyzed.

[0014] Preferably, the mapping relationship conflict amount of each hash key value is determined by the product result of the number of data types and the data chaos.

[0015] Preferably, the data chaos is further determined by the information entropy of all data under the same hash key value.

[0016] Preferably, the calculation method of the arrangement priority level of each hash key value is as follows:

[0017] Calculate the product of the probability of the data with the highest frequency of occurrence among all data for each hash key value and the maximum value of the frequency of occurrence of each hash key value among all its data types.

[0018] Take the ratio result of the product to the mapping relationship conflict amount as the arrangement priority level of each hash key value.

[0019] Preferably, the method of screening out a preset number of hash key values is further determined as screening out the hash key values corresponding to the first preset number of arrangement priority levels in descending order of the arrangement priority levels of all hash key values.

[0020] Preferably, the priority level update amount of the hash key values in the cache data table is determined by the product result of the average levels of the two difference values after differentiation in step A3.

[0021] Preferably, the average level of the difference values after differentiation is further determined by the average value of the absolute values of all the difference values after differentiation.

[0022] Preferably, the method for updating the priority level of the arrangement of the hash key values in the cache data table is as follows:

[0023] For each hash key value in the cache data table, when the data to be analyzed scheduled is the data corresponding to the hash key value, the priority level of the arrangement of this hash key value is added with the priority level update amount; otherwise, the priority level of the arrangement of this hash key value is subtracted by the priority level update amount.

[0024] Preferably, the method for updating the cache data table in the GPU shared memory is as follows:

[0025] When the hash key value of the data to be analyzed is in the cache data table, all the hash key values in the cache data table are re-sorted from largest to smallest according to the updated priority level of the arrangement;

[0026] Otherwise, the priority level of the arrangement of the hash key value of the data to be analyzed is used to determine whether the corresponding hash key value is added to the cache data table.

[0027] Preferably, the method for determining whether the corresponding hash key value is added to the cache data table is as follows:

[0028] When the priority level of the arrangement of the hash key value of the data to be analyzed is greater than or equal to the priority level of the arrangement of the last hash key value in the cache data table, the hash key value of the data to be analyzed replaces the last hash key value in the cache data table;

[0029] Otherwise, no adjustment is made to the cache data table.

[0030] This application has at least the following beneficial effects:

[0031] Using a FIFO queue to manage cached data in GPU shared memory in traditional methods results in low-frequency data also occupying cache space, which affects data scheduling efficiency. This application optimizes the scheduling of nose data and GPU parallel processing. By obtaining the data to be analyzed in the nose data in real time, analyzing the number of data types and the chaos of data that appear under the same hash key value for all data, calculating the mapping relationship conflict amount of each hash key value to reflect the conflict amount between data; combining the mapping relationship conflict amount, the probability of the data with the most frequent occurrences among all data for each hash key value, and the maximum value of the frequencies of occurrences of each hash key value among all its data types, calculating the permutation priority level of each hash key value to reflect the priority level of data; analyzing the average level of the difference values after differentiating all hash key values according to the time sequence of entry into the cache data table in the real-time cache data table, and the average level of the difference values after differentiating according to the permutation priority level of all hash key values, calculating the priority level update amount of the hash key values in the cache data table, and using the priority level update amount to update the permutation priority level of the hash key values in the cache data table, that is, dynamically adjusting the permutation priority level of the hash key values in combination with the real-time data scheduling situation, ensuring that the permutation priority level of the hash key values changes continuously with the GPU running time. This application optimizes the scheduling of nose data and GPU parallel processing, avoids the problem of low data scheduling efficiency in the prior art, and can effectively improve data processing efficiency and GPU running efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a flowchart of a method for optimizing the scheduling of nose data and GPU parallel processing provided by the present application;

[0034] Figure 2 It is a flowchart of the process for updating the permutation priority level of hash key values in the cache data table of GPU shared memory provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] To further elaborate on the technical means and effects adopted by this application to achieve the intended invention purpose, the following will, in conjunction with the accompanying drawings and preferred embodiments, elaborate in detail on a scheduling optimization method for parallel processing of nose data and GPU, its specific implementation manner, structure, features, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs.

[0037] The following will specifically describe the specific solution of a scheduling optimization method for parallel processing of nose data and GPU provided by this application in conjunction with the accompanying drawings.

[0038] A scheduling optimization method for parallel processing of nose data and GPU provided by an embodiment of this application.

[0039] Specifically, a scheduling optimization method for parallel processing of nose data and GPU is provided as follows. Please refer to Figure 1 , and this method includes the following steps:

[0040] Step 1: Obtain all the data in the GPU global memory and the data to be analyzed in the nose command;

[0041] Before the GPU of the server runs a computing task, the server receives various request commands issued by the user through the nose, obtains the nose data, and transfers the nose data to the CPU of the server. The CPU analyzes the commands in the nose data and transfers the data that requires GPU participation in the calculation to the global memory of the GPU. The global memory is the main memory space of the GPU and can be accessed by all cores.

[0042] Next, the GPU will process these data in chunks. In NVIDIA GPUs, the basic computing unit is the Streaming Multiprocessor (SM for short), and each SM can execute multiple threads in parallel. When the threads schedule data, there is a shared memory in the GPU for caching data to facilitate the rapid scheduling of threads in the SM. After the GPU calculation is completed, the result of the GPU calculation is transmitted to the CPU, and the CPU will package the nose data and the CPU calculation result into response data and transmit it to the user through the network.

[0043] Thus, all the data in the GPU global memory and the data to be analyzed in the nose command can be obtained.

[0044] Step 2: Update the priority level of the hash key values in the cache data table of the GPU shared memory according to the distribution of the hash key values of all data and the real-time access situation in the cache data table.

[0045] In this application, the process flow chart for updating the priority level of the hash key values in the cache data table of the GPU shared memory is as shown in the appendix Figure 2 as follows:

[0046] A1. Calculate the mapping relationship conflict amount of each hash key value by analyzing the number of data types and the chaos of data that appear under the same hash key value for all data.

[0047] For the shared memory of the GPU, since the shared memory of the GPU is relatively small and the amount of data that can be stored is relatively small, in this application, a hash table scheduling hash function is used to perform a hash operation on the data in the global memory to obtain the hash key values of the data. When storing the hash key values of the data in the shared memory of the GPU and performing a search, a mapping search is performed through the hash key values. Among them, the hash function operation of the data and the search for data through the hash key values are well-known technologies, and the specific calculation process will not be elaborated here.

[0048] After mapping all the data in the global memory through the hash function, the hash key values may have hash conflicts due to hash mapping, so that different data may be mapped to the same hash key value. Therefore, if a hash key value corresponds to multiple data, more time is required to search for data through this hash key value, which reduces the search effect. Therefore, when arranging this hash key value in the GPU shared memory, it should be arranged at the end of the cache data table so that other data can be searched in advance.

[0049] Therefore, calculate the mapping relationship conflict amount of each hash key value. Taking the mapping relationship conflict amount of the i-th hash key value in the global memory as an example, the expression is:

[0050] ;

[0051] In the formula, represents the mapping relationship conflict amount of the i-th hash key value; represents the number of data types corresponding to the i-th hash key value; represents the information entropy of all data corresponding to the i-th hash key value. Among them, the calculation of the information entropy is a well-known technology, and the specific calculation process will not be elaborated here. It is used to characterize the chaos of all data corresponding to the i-th hash key value, and variance can also be used to characterize it in other embodiments.

[0052] For the hash key values of data in the global memory of a GPU, when the types of data corresponding to the hash key values are more, that is the larger the value is, it indicates that when the data is mapped to the hash key value, the possibility of conflict is greater. In the process of the GPU thread scheduling data, the query time required for scheduling through this hash key value is longer compared to the hash key value corresponding to fewer data types; when the frequencies of the data corresponding to the hash key value are more chaotic, it means that for different data corresponding to this hash key value, the scheduling frequencies of each data are more uneven. Therefore, the value of the information entropy of the frequencies of the data corresponding to this hash key value is larger.

[0053] Thus, during the arrangement process of this hash key value, it should be arranged at the back of the cache data table so that the data that can be quickly queried can be scheduled and queried first, reducing the query scheduling time of the data.

[0054] In other embodiments of the present application, the mapping relationship conflict amount of the i-th hash key value can also be recorded as: .

[0055] A2. Combining the mapping relationship conflict amount, the probability of the data with the most occurrences of each hash key value among all data, and the maximum value of the frequencies of each hash key value among all its data types, calculate the arrangement priority level of each hash key value; screen out a preset number of hash key values according to the size of the arrangement priority level and store them in the cache data table of the GPU shared memory;

[0056] During the process of searching for data through the hash key value, the threads of the GPU need to first traverse the cache data table in the shared memory. If the frequency of the data appearance is higher, it means that the data is scheduled by the threads in the GPU more times. Arranging its hash key value in the front of the cache data table can effectively reduce the thread scheduling data time. And when the hash key value corresponds to multiple data, the data with the highest frequency of occurrence is given priority, which can reduce the conflicts during traversal.

[0057] Therefore, by analyzing the frequencies of data appearance, the arrangement priority level of each hash key value can be calculated. Taking the arrangement priority level of the k-th hash key value in the global memory as an example, the expression is:

[0058] ;

[0059] In the formula, represents the arrangement priority level of the k-th hash key value; represents the probability of the data type with the most occurrences of the k-th hash key value among all data; represents the maximum value of the frequencies of the k-th hash key value among all its data types; Represents the conflict volume of the mapping relationship of the k-th hash key value. The calculation of data probability is a well-known technology, and the specific calculation process will not be elaborated here.

[0060] For the multi-threaded parallel computing of the GPU, different threads may adopt the same data. Therefore, the number of scheduling times for different data is different. The more times the data is scheduled by the GPU threads, the higher the utilization rate of the data, and the higher the probability of scheduling this data in the future. For the data corresponding to the hash key value, when the number of data types corresponding to the hash key value is less, the time required for scheduling query is less. Therefore, the maximum probability value of the data Is used as the scheduling probability weight; therefore, the frequency of occurrence of this data Is used as the possibility. When both values are larger, it indicates that the possibility of future scheduling of the data corresponding to this hash key value is higher. At the same time, the conflict volume of the mapping relationship of the hash key value Is smaller, and the time for scheduling the data corresponding to this hash key value is less, making the priority level of the arrangement of this hash key value.

[0061] Then, all hash key values are arranged in descending order according to their priority levels, and the first N hash key values after arrangement are selected. N takes the value of 32 in this embodiment and can be set according to the actual situation. The selected hash key values are stored in the cache data table in the shared memory of the GPU.

[0062] A3. Analyze the average level of the difference values after differentiating all hash key values in the real-time cache data table according to the time sequence of entering the cache data table, and the average level of the difference values after differentiating according to the priority levels of all hash key values, and calculate the priority level update amount of the hash key values in the cache data table;

[0063] During the multi-threaded operation of the GPU, it may occur that at the initial state of the thread, a large number of threads need the same data, resulting in a relatively large number of scheduling times for this data. Therefore, the priority level of the hash key value corresponding to this data is relatively large. However, as the program executes, some data may no longer be frequently accessed, but their corresponding hash key values may still be retained in the cache, resulting in repeated traversal of this hash key value during the subsequent thread scheduling data process, increasing the time for thread scheduling data. Therefore, when the GPU is running in real time, it is necessary to update the cache data table in the shared memory according to the scheduling time of the data.

[0064] For the cache data table in the shared memory of the GPU, in the best case, the order of the hash key values conforms to the scheduling time, that is, the closer to the current moment, the more forward the hash key values are arranged, which can save the scheduling time.

[0065] Thus, for the hash key values in the cache data table, perform a first-order difference calculation according to the time sequence of the hash key values stored in the cache data table, and then calculate the mean value of the absolute values of all difference values to obtain the scheduling priority time difference degree, denoted as Q, which is used to characterize the degree of change in the data scheduling type. Among them, the calculation of the first-order difference method is a well-known technology, and the specific calculation process will not be elaborated here. Based on the above analysis, calculate the priority level update amount V of the hash key values in the cache data table, and the expression is:

[0066] ;

[0067] In the formula, represents the priority level update amount of the hash key values in the cache data table; represents the scheduling priority time difference degree of the hash key values; represents the class difference degree of the hash key values in the cache data table, which is used to characterize the degree of change in data scheduling. The calculation method of the class difference degree of the hash key values in the cache data table is: perform a first-order difference on the arrangement priority degree of the hash key values in the cache data table and then calculate the mean value of the absolute values of all difference values.

[0068] It should be understood that when the data is continuously called, the higher the priority of the data scheduled by the GPU thread, the greater the urgency of the data being utilized, and the higher the priority of scheduling this data in the future. The more forward the arrangement order of the hash key values in the cache table, the less the change in the data scheduling type. Therefore, the value of the scheduling priority time difference degree Q of the hash key values is smaller, and the value of the class difference degree R of the hash key values is smaller, ultimately resulting in a smaller value of the priority level update amount V of the hash key values. When there are significant changes in data scheduling, the scheduling time of the hash key values in the cache data table will be disrupted. Therefore, it is necessary to significantly increase the arrangement priority degree of the hash key values of the currently scheduled data, that is, the value of the scheduling priority time difference degree is larger, so that the value of the priority level update amount of the hash key values in the cache data table is larger, thereby improving the efficiency of the GPU multi-threaded scheduling data.

[0069] In other embodiments of the present application, the priority level update amount V of the hash key values in the cache data table can also be set to .

[0070] A4, and use the priority level update amount to update the arrangement priority degree of the hash key values in the cache data table;

[0071] Based on the above steps, update the arrangement priority degree of the hash key values through the real-time scheduled data. Taking the updated arrangement priority degree of the x-th hash key value in the cache data table as an example, the update method is:

[0072] ;

[0073] In the formula, represents the updated permutation priority level of the x-th hash key value in the cache data table; represents the permutation priority level of the x-th hash key value in the cache data table before update; represents a scheduling tag, where c = 1 indicates that the data to be analyzed for scheduling is the data corresponding to this hash key value, and c = 0 indicates that the data to be analyzed for scheduling is not the data corresponding to this hash key value.

[0074] In the multi-threaded operation of the GPU, the frequent scheduling of data has a direct impact on the permutation priority of hash key values. When a certain data is frequently scheduled recently, the permutation priority of its corresponding hash key value should be increased, that is , so as to keep its high-priority state in the cache for quick access. On the contrary, for those data that are no longer frequently scheduled, the permutation priority of their hash key values should be decreased, that is , so that the sorting of these data in the cache data table in the GPU shared memory gradually decreases.

[0075] This dynamic adjustment helps to optimize the use of the cache, reduce the caching of data that is no longer needed, and thus make room for data that is accessed more frequently. As the data scheduling pattern changes, the GPU needs to update the cache data table in the shared memory in real time to reflect the latest data access pattern. This update can be achieved by monitoring the scheduling time of the data, thereby adjusting the permutation priority of the hash key values. When the data is frequently scheduled, increasing the priority of its hash key value can reduce the time for future GPU threads to schedule data through the shared memory and improve the operating efficiency of the GPU.

[0076] Step 3: Update the cache data table in the GPU shared memory by using the permutation priority level of the hash key value of the data to be analyzed.

[0077] Update the cache data table in the GPU shared memory through the updated permutation priority level of the hash key value at any time as follows:

[0078] When the hash key value of the data to be analyzed is in the cache data table, re-sort all the hash key values in the cache data table from largest to smallest according to the updated permutation priority level;

[0079] Otherwise, use the permutation priority level of the hash key value of the data to be analyzed to determine whether its corresponding hash key value is added to the cache data table.

[0080] Among them, the method for judging whether its corresponding hash key value is added to the cache data table is:

[0081] When the permutation priority degree of the hash key value of the data to be analyzed is greater than or equal to the permutation priority degree of the last hash key value in the cache data table, replace the last hash key value in the cache data table with the hash key value of the data to be analyzed;

[0082] Conversely, no adjustment is made to the cache data table.

[0083] This kind of dynamic adjustment helps to reduce conflicts in the cache and improve the efficiency of data retrieval. By adjusting the priority of the hash key value in real time, the performance degradation caused by conflicts can be reduced, ensuring that the GPU can quickly respond to data access requests, reducing the time for thread scheduling data in parallel, and improving the operating efficiency of the GPU. After each update of the cache data table, the multi-threads in the GPU will use this updated cache data table to schedule data, and retrieve and use data according to the latest priority order, thereby improving the data retrieval efficiency.

[0084] Each embodiment in this application is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

[0085] It should be noted that unless otherwise specified and limited, terms such as "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the article or device including the element. In addition, the term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0086] Those skilled in the art will readily think of other implementation schemes of this application after considering the specification and practicing the invention herein. This application is intended to cover any variations, uses or adaptations of this application, which follow the general principles of this application and include common general knowledge or conventional technical means in the technical field not invented by this application.

[0087] It should be understood that this application is not limited to the exact structure already described and shown in the drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A scheduling optimization method for head data and GPU parallel processing, characterized in that: The method comprises the following steps: Step 1: Get all the data in the GPU global memory and the data to be analyzed in the head command; Step 2: Update the arrangement priority of the hash key values ​​in the cache data table of the GPU shared memory according to the distribution of the hash key values ​​of all data in the GPU global memory and the real-time access in the cache data table; specifically: A1, by analyzing the number of data types and data confusion under the same hash key value in all data in the GPU global memory, calculate the mapping conflict amount of each hash key value; A2, combining the mapping relationship conflict amount, the probability of the data with the highest frequency of each hash key value in all data in the GPU global memory, and the maximum frequency of each hash key value in all data types in its GPU global memory, calculate the arrangement priority of each hash key value; select a preset number of hash key values ​​according to the size of the arrangement priority, and store them in the cache data table of the GPU shared memory; A3, analyze the average level of the difference values ​​of all hash key values ​​in the real-time cache data table after being differentiated according to the time sequence of entering the cache data table, and the average level of the difference values ​​after being differentiated according to the arrangement priority of all hash key values, and calculate the priority update amount of the hash key values ​​in the cache data table; A4, using the priority update amount, updates the arrangement priority of the hash key values ​​in the cache data table; Step 3: When the hash key value of the data to be analyzed is in the cache data table, all hash key values ​​in the cache data table are reordered from large to small according to the updated arrangement priority; otherwise, the arrangement priority of the hash key value of the data to be analyzed is used to determine whether its corresponding hash key value is added to the cache data table.

2. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The mapping relationship conflict amount of each hash key value is determined by the product of the number of data types that appear and the disorder of the data.

3. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The disorder of the data is further determined by the information entropy of all data in the GPU global memory under the same hash key value.

4. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The calculation method of the arrangement priority of each hash key value is: Calculate the product of the probability of each hash key value appearing most frequently in all data in the GPU global memory and the maximum value of the frequency of each hash key value appearing in all data types in its GPU global memory; The ratio of the product to the conflict amount of the mapping relationship is used as the arrangement priority level of each hash key value.

5. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The method for screening out a preset number of hash key values ​​is further determined as follows: arranging all hash key values ​​in descending order of priority, and screening out hash key values ​​corresponding to the first preset number of arrangement priorities.

6. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The priority update amount of the hash key value in the cache data table is determined by the product of the average levels of the two differential values ​​in step A3.

7. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 6, characterized in that: The average level of the post-differentiation difference values ​​is further determined by the average of the absolute values ​​of all the post-differentiation difference values.

8. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The update method of the arrangement priority of the hash key values ​​in the cache data table is: For each hash key value in the cache data table, when the scheduled data to be analyzed is the data corresponding to the hash key value, the arrangement priority of the hash key value is added to the priority update amount; otherwise, the arrangement priority of the hash key value is subtracted from the priority update amount.

9. The scheduling optimization method for head data and GPU parallel processing as claimed in claim 1, characterized in that: The method for determining whether the corresponding hash key value is added to the cache data table is: When the arrangement priority of the hash key value of the data to be analyzed is greater than or equal to the arrangement priority of the last hash key value in the cache data table, the hash key value of the data to be analyzed replaces the last hash key value in the cache data table; Otherwise, no adjustment is made to the cache data table.

Images