Wafer dispatch strategy query method and device based on intelligent parallel data engine

Through the intelligent parallel data engine's dispatch strategy query method, the problem of slow reading speed of real-time dispatch system in massive data processing is solved, the dispatch efficiency and system stability are improved, and the wafer factory's production capacity is maximized .

CN117251476BActive Publication Date: 2024-11-08上海朋熙半导体股份有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311372089.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-10-23
Publication Date
2024-11-08
Estimated Expiration
2043-10-23

AI Technical Summary

Technical Problem

The existing real-time dispatch system is slow to read when processing massive data, which is prone to stuttering, resulting in low efficiency of dispatch and inability to provide optimal dispatch services, limiting the improvement of the production capacity of the wafer factory.

Method used

Using the wafer dispatch strategy query method of intelligent parallel data engine, the node dependency graph is generated by analyzing the dispatch strategy to be queried, and tasks are processed in batches according to the dependency level, and IO and CPU tasks are executed in parallel using thread pools. , ensure the improvement of data reading efficiency.

Benefits of technology

It significantly improves the data reading capability of the dispatch system, ensures real-time and accurate dispatch, improves system stability and dispatch efficiency, and maximizes production capacity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117251476B_ABST
    Figure CN117251476B_ABST
Patent Text Reader

Abstract

The present application provides a wafer dispatching strategy query method and device for an intelligent parallel data engine, proposes an intelligent parallel data engine, analyzes the fields used according to the scheduling rules and generates a node dependency graph according to their relationship, further configures aggregation and generates tasks according to the dependency relationship, and obtains the same type of tasks at one time through parallel query, divides the tasks into IO-type tasks and CPU-type tasks, and executes them in parallel using a thread pool. The application of the present invention will greatly improve the data reading efficiency of the dispatching system, and solve the problems of slow reading speed, easy jamming, and unstable performance of the current dispatching system when facing massive data. The improvement of data reading capability can ensure that the dispatching system can further realize real-time and accurate dispatching, improve system stability, optimize the quality of the dispatching system, and ultimately improve dispatching efficiency and maximize production capacity.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of methods and equipment for manufacturing semiconductor devices, and in particular to a method and equipment for querying wafer dispatching strategies based on an intelligent parallel data engine. Background Art

[0002] Chip manufacturing is known for its complex process, which requires hundreds of steps before it can be manufactured. During production, there are many constraints such as specific machine restrictions, time restrictions between processes, and upstream and downstream balance. In addition, there are often hundreds of products with different delivery orders and tens of thousands of wafer batches waiting to be produced in the factory.

[0003] Faced with this, chip factories often use real-time dispatching systems (RTD) to implement scheduling and dispatching. The real-time means that the dispatching system needs to control the factory in all aspects and details, and interact with the on-site MES, MCS, EAP and other systems in real time. As a result, massive amounts of data are generated. The advent of the Industrial 4.0 era has further increased the amount of data in wafer fabs. The real-time dispatching system needs to process up to thousands of data per second. The data reading capability will affect the system's response speed and is one of the key indicators for judging the performance of the real-time dispatching system.

[0004] However, the current real-time dispatching system is not satisfactory in data reading, is prone to jamming, and cannot provide the best dispatching service, which in turn restricts the further improvement of wafer fab production capacity. How to improve the data reading capability of the current real-time dispatching system is of great significance to wafer fabs. Summary of the invention

[0005] One purpose of the present application is to provide a method and device for querying wafer dispatching strategies of an intelligent parallel data engine, at least to solve the technical problem of low dispatching efficiency caused by slow data reading and unstable performance.

[0006] To achieve the above objectives, some embodiments of the present application provide the following aspects:

[0007] In a first aspect, the present application provides a wafer dispatch strategy query method of an intelligent parallel data engine, comprising:

[0008] Parse the wafer dispatching strategy to be queried and obtain the fields required by the dispatching strategy;

[0009] Taking the fields as nodes, generating a node dependency graph according to the dependency relationship between the fields;

[0010] Divide the nodes into batches according to the dependency levels between fields, and generate and execute tasks for each batch of nodes in order from shallow to deep levels until all batch nodes are processed;

[0011] Feedback the final query results.

[0012] Optionally, the parsing of the wafer dispatching strategy to be queried to obtain fields required by the dispatching strategy includes:

[0013] Obtain at least one model corresponding to the wafer dispatching strategy to be queried;

[0014] Required fields are extracted from the at least one model according to the content of the work dispatching strategy.

[0015] Optionally, after generating the node dependency graph according to the dependency relationship between the fields, the following is further included:

[0016] Determine whether the node dependency graph is a directed acyclic graph;

[0017] If it is a directed acyclic graph, continue to the next step. If it is not a directed acyclic graph, an error is reported.

[0018] Optionally, the method further includes:

[0019] If it is a directed acyclic graph, mark all nodes as unprepared;

[0020] The tasks are generated and executed according to each batch of nodes in the order of the levels from shallow to deep, including:

[0021] Query each node in this batch;

[0022] For any node in this batch, if the status of the upper-level dependent nodes of the node are all in the ready state, generate a task and execute it; update the status of each node in this batch to the ready state;

[0023] Determine whether all nodes are in the ready state; if not, query the next batch of nodes and return the operation of generating tasks and executing them; if yes, execute the operation of feeding back the final query results;

[0024] If some tasks in this batch fail to execute, the failed tasks will be regenerated and executed during the next batch processing.

[0025] Optionally, query the next batch of nodes and return the generated tasks and perform the following operations:

[0026] Determine whether the next batch reaches the set threshold;

[0027] If yes, report an error; if not, return the action that generates the task and executes it.

[0028] Optionally, the generating and executing of the task includes:

[0029] Divide tasks into IO-type tasks and CPU-type tasks for execution;

[0030] Tasks in the same batch have no dependencies on each other and are executed in parallel.

[0031] Optionally, the IO-type tasks are executed in parallel using a first thread pool, and the CPU-type tasks are executed in parallel using a second thread pool; the number of threads in the first thread pool is greater than that in the second thread pool.

[0032] Optionally, the generating and executing of the task includes:

[0033] For tasks of the same type, the same task can be generated and executed by obtaining nodes through a single query.

[0034] Optionally, after generating and executing tasks according to each batch of nodes in the order of hierarchy from shallow to deep, the following is also included:

[0035] Write the execution result back to the context.

[0036] In a second aspect, the present application provides an electronic device, including: a memory and a processor;

[0037] Among them, one or more computer programs are stored in the memory, and the one or more computer programs include instructions; when the instructions are executed by the processor, the electronic device executes any wafer dispatch strategy query method of the intelligent parallel data engine.

[0038] Compared with the prior art, the present invention provides an optimization solution for intelligent scheduling - an intelligent parallel data engine, which analyzes the fields used according to the scheduling rules and generates a node dependency graph according to their relationships, and further configures aggregation and generates tasks according to the dependency relationships. Tasks of the same type are obtained once through parallel queries, and tasks are divided into IO-type tasks and CPU-type tasks, which are executed in parallel using thread pools. The application of the present invention will greatly improve the data reading efficiency of the dispatching system, and solve the problems of slow reading speed, easy jamming, and unstable performance of the current dispatching system when facing massive data. The improvement of data reading capabilities can ensure that the dispatching system can further realize real-time and accurate dispatching, improve system stability, optimize the quality of the dispatching system, and ultimately improve dispatching efficiency and maximize production capacity. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flowchart of a wafer dispatch strategy query method of an intelligent parallel data engine provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of the fields included in the work dispatching strategy provided in the embodiment of the present application;

[0041] Figure 3 A DAG diagram of the work dispatching strategy provided in the embodiment of the present application;

[0042] Figure 4A schematic diagram of a data reading process based on an intelligent parallel data engine provided in an embodiment of the present application;

[0043] Figure 5 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0045] Embodiment 1

[0046] The present application provides a wafer dispatch strategy query method based on an intelligent parallel data engine, which is applicable to the case where dispatch strategy data is read based on an intelligent parallel data engine in a wafer production scenario. Figure 1 , the method comprising:

[0047] S11. Parse the wafer dispatching strategy to be queried to obtain fields required by the dispatching strategy.

[0048] The user inputs the wafer dispatching strategy as a query condition into the intelligent parallel data engine, and the intelligent parallel data engine parses the strategy. Optionally, at least one model corresponding to the wafer dispatching strategy to be queried is obtained; and the required fields are extracted from the at least one model according to the content of the dispatching strategy.

[0049] Taking Strategy 001 as an example, the strategy goal is to determine the order of wafer dispatch according to the rules used in the strategy. Figure 2 , it is set that all the fields used in this strategy are derived from the Mes database. The fields used in this strategy come from three models: wafer model, process step model and machine model. Each model includes a field table, and the primary key PK in the field table is a specific field.

[0050] It should be noted that the dispatch strategy Strategy 001 is only used as an example to illustrate the fields required for the dispatch strategy and does not represent the dispatch rules actually used in the wafer fab.

[0051] According to the content of the dispatching strategy, the required fields are extracted from the wafer model, the process step model and the machine model, as follows:

[0052] (1) Wafer model, referred to as Lot model: Lot and wafer batch. The Lot model contains the relevant attribute information of the wafer batch. In the set Strategy 001, the Lot model contains the following fields: lotId is the wafer batch, lotPriority is the wafer priority, currentStepId is the ID of the process step where the wafer batch is located, step is the process step where the wafer batch is located, dueDate is the delivery date of the wafer batch, remaningQtime refers to the remaining waiting time, remaningCycleTime refers to the production time required for the wafer to be completed, scheduleScore is the scheduling score of the wafer batch, which is calculated and counted by the Mes system, criticalRatio is the ratio of the remaining production time and the remaining delivery time of the wafer batch, which needs to be calculated through dueDate and remaningCycleTime. isSameReticle indicates whether the photomask required for the production of the wafer batch is consistent with the photomask currently loaded on the dispatching machine.

[0053] (2) Process step model, referred to as Step model: contains all the Step information of each product. It is simplified here and only shows the three fields used by Strategy 001. stepId is the process step ID to which the wafer batch belongs, step is the process step of the wafer batch, and currentReticleId is the photomask ID required for the current step of the wafer batch.

[0054] (3) Equipment model, referred to as Eqp model: The Eqp model contains the relevant attributes of the equipment. It is simplified here and only shows the fields used by Strategy 001, including equipmentId, i.e., the equipment ID, and the photomask ID currently loaded on the machine, i.e., currentReticleId.

[0055] S12: Taking the fields as nodes, generating a node dependency graph according to the dependency relationship between the fields.

[0056] Each field obtained in S11 is taken as a node and a node element of the node dependency graph.

[0057] There are dependencies between fields, not limited to the same model or between models. For example, isSameReticle in the Lot model needs to be associated with the step model according to the currentStepID of the Lot, and the photomask ID (currentReticleId) used in the current process step of the wafer lot is found by comparing the photomask ID currently required by the wafer lot with the photomask ID loaded on the machine. For other dependencies, see Figure 2 The dashed arrow in .

[0058] Transform the dependency relationship into a one-way arrow between nodes to form a node dependency graph. Furthermore, the node dependency graph should be a directed acyclic graph, that is, no circular dependencies are allowed between nodes, see Figure 3 , otherwise an error is reported.

[0059] S13. Divide the nodes into batches according to the dependency levels between fields, generate tasks for the nodes in each batch and execute them in order from shallow to deep levels until all batch nodes are processed.

[0060] See also Figure 3 , lotId and equipmentId do not depend on any node, and are nodes with known status, which is the first batch Batch0. lotPriority, currentStepId, etc. depend on lotId, that is, the priority of the wafer and the ID of the process step in which the wafer batch is located can be obtained only after knowing the specific wafer batch, which is the second batch Batch1. criticalRatio needs to be calculated by dueDate and remaningCycleTime, which is the third batch Batch2. isSameReticle needs to associate the step model according to the currentStepID of Lot, find the photomask ID (currentReticleId) used in the current process step of the wafer batch, and compare the photomask ID currently required by the wafer batch with the photomask ID loaded on the machine, which is the fourth batch Batch3. Mark the dependency level and batch of each node.

[0061] In order to facilitate the management of whether each node has a query and whether it is computable, all nodes need to be marked as unready before generating tasks, and tasks are generated and executed according to the order of the levels from shallow to deep according to the nodes of each batch. For each batch, first query each node of this batch. For any node in this batch, if the status of the upper-level dependent nodes of the node is ready, the node status is computable, and tasks are generated and executed. Update the status of each node in this batch to ready.

[0062] Determine whether all nodes are in the ready state; if not, query the next batch of nodes and return the operation of generating tasks and executing them; if yes, execute the operation of feeding back the final query results. And so on, until all batches of nodes are processed. In this way, you can query layer by layer according to the dependency level between fields, improving query efficiency.

[0063] If some tasks in this batch fail to execute, the failed tasks will be regenerated and executed during the next batch processing.

[0064] Furthermore, a batch threshold is set to avoid errors caused by too many levels or loops. Specifically, the operations of querying the nodes of the next batch and returning to generate tasks and execute them include: determining whether the next batch reaches the set threshold; if so, reporting an error; if not, returning to generate tasks and execute them.

[0065] S14. Feedback the final query result.

[0066] After executing the task, the nodes of the previous batch write the execution results back to the context, which serves as the initial value of the nodes of the next batch. The query result messages of all the last-level nodes are assembled and fed back to the requester.

[0067] Embodiment 2

[0068] This embodiment refines "generating and executing tasks" based on the above embodiment. Specifically, tasks are divided into IO-type tasks and CPU-type tasks for execution.

[0069] Specifically, IO tasks are tasks that mainly occupy IO resources, such as querying external interfaces and accessing databases during policy execution. CPU tasks are such as expression calculation and function execution. IO tasks are executed in parallel using the first thread pool, and CPU tasks are executed in parallel using the second thread pool; the number of threads in the first thread pool is greater than that in the second thread pool.

[0070] Optionally, in tasks of the same type, the same task can be generated and executed by obtaining the nodes through a single query. For example, if the fields corresponding to at least two nodes in a batch come from the same table, the fields of at least two nodes can be obtained through a single query, and the same task can be generated. Figure 3 In Batch1, the field query in the lot model is defined as a task task1, and the query of the field in the Eqp model is defined as a task task2.

[0071] Optionally, tasks in the same batch have no dependencies on each other and can be executed in parallel. Figure 3 , task1 and task2 of Batch1 are IO tasks, which can be executed in parallel with a larger thread pool. criticalRatio in Batch2 and isSameReticle in Batch3 are CPU tasks. When the dispatching strategy is more complex and there are many tasks to be executed, a smaller thread pool can be selected for parallel execution.

[0072] This application can improve the efficiency of task generation and execution through the above-mentioned same task generation strategy and task parallel strategy.

[0073] Embodiment 3

[0074] In one specific embodiment, using Figure 4 The process shown executes the method provided by this application.

[0075] Step 1: Analyze the query conditions. The query conditions are specific work dispatch strategies.

[0076] Step 2: Generate a node dependency graph based on the dependency relationship. Different BatchIds are marked on the dependency graph.

[0077] Step 3: Determine whether the node dependency graph is a DAG graph. If not, return an exception. If yes, proceed to step 4.

[0078] Figure 3 The node dependency graph generated after parsing the dispatch strategy Strategy 001 is shown. Each field in the three data models used by the strategy corresponds to a node. All query targets, i.e. all fields used by the strategy, are added to the graph. Then criticalRatio depends on dueDate and remaningCycleTime for calculation, and isSameReticle depends on currentStepId and step in the Lot model; stepId and currentReticleId in the Step model; and currentReticleId in the Eqp model. The node dependency graph generated by Strategy 001 is a DAG graph after adding its dependencies to the graph.

[0079] Step 4: The node status is marked as unready. Figure 4 , the initial state is that lotId and euipmenId are known as Batch0. In the subsequent basic model, data that does not need to depend on other nodes is divided into Batch1, data that depends on other nodes in the first layer is divided into Batch2, and data that depends on other nodes in the second layer is divided into Batch3. If there is a problem with data reading, the unread fields will be automatically divided into batchId+1 and continue to execute.

[0080] Step 5: Initialize BatchId = 0. Initially, all nodes used are in non-computable state (NONCOMPUTABLE), BatchId = 0, and start querying

[0081] Step 6: Determine whether BatchId is less than the threshold. If so, proceed to step 7; otherwise, return an exception.

[0082] Step 7: Generate tasks based on computable nodes.

[0083] Step 8: Execution of IO-type tasks and CPU-type tasks. For the specific process of generating tasks and executing tasks, please refer to the description of the second embodiment.

[0084] Step 9: Write the execution result back to the context.

[0085] Step 10: Update the node status of this batch to Ready.

[0086] Step 11: Determine whether all nodes are in the Ready state. If not, add BatchId++ and return to step 6; if yes, proceed to step 12. Figure 4 , Strategy 001 needs to be performed at least four times to turn all nodes into READY state.

[0087] Step 12: Assemble the query result message and return the execution result to the requester.

[0088] Embodiment 4

[0089] Figure 5 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device 500 includes one or more processors 501 and a memory 502.

[0090] The processor 501 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 500 to perform desired functions.

[0091] The memory 502 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 501 may run the program instructions to implement the wafer dispatch strategy query method of an intelligent parallel data engine of any embodiment of the present application described above and / or other desired functions. Various contents such as initial external parameters, thresholds, etc. may also be stored in the computer-readable storage medium.

[0092] In one example, the electronic device 500 may further include: an input device 503 and an output device 504, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown). The input device 503 may include, for example, a keyboard, a mouse, etc. The output device 504 may output various information to the outside, including early warning information, braking force, etc. The output device 504 may include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, etc.

[0093] Of course, to simplify, Figure 5 Only some of the components related to the present application in the electronic device 500 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device 500 may also include any other appropriate components.

[0094] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of a wafer dispatch strategy query method for an intelligent parallel data engine provided in any embodiment of the present application.

[0095] The flow chart or block diagram in the accompanying drawings shows the possible architecture, function and operation of the equipment, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated system for hardware that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0096] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the present application. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A wafer dispatch strategy query method for an intelligent parallel data engine, characterized in that: include: Parse the wafer dispatching strategy to be queried and obtain the fields required by the dispatching strategy; Taking the fields as nodes, generating a node dependency graph according to the dependency relationship between the fields; Divide the nodes into batches according to the dependency levels between fields, and generate and execute tasks for each batch of nodes in order from shallow to deep levels until all batch nodes are processed; Feedback the final query results; The generating and executing of tasks includes: dividing the tasks into IO type tasks and CPU type tasks for execution; tasks in the same batch have no dependency on each other and are executed in parallel; The generating and executing of tasks includes: in tasks of the same type, the same task can be generated and executed by nodes obtained through one query.

2. The method according to claim 1, characterized in that The parsing of the wafer dispatching strategy to be queried to obtain the fields required by the dispatching strategy includes: Obtain at least one model corresponding to the wafer dispatching strategy to be queried; Required fields are extracted from the at least one model according to the content of the work dispatching strategy.

3. The method according to claim 1, characterized in that: After generating the node dependency graph based on the dependencies between fields, it also includes: Determine whether the node dependency graph is a directed acyclic graph; If it is a directed acyclic graph, continue to the next step. If it is not a directed acyclic graph, an error is reported.

4. The method according to claim 3, characterized in that The method further comprises: if it is a directed acyclic graph, marking all nodes as unprepared; The generating and executing of tasks according to the nodes of each batch in the order of the levels from shallow to deep includes: querying each node of the batch; For any node in this batch, if the status of the upper-level dependent nodes of the node are all in the ready state, generate a task and execute it; update the status of each node in this batch to the ready state; Determine whether all nodes are in the ready state; if not, query the next batch of nodes and return the operation of generating tasks and executing them; if yes, execute the operation of feeding back the final query results; If some tasks in this batch fail to execute, the failed tasks will be regenerated and executed during the next batch processing.

5. The method according to claim 4, characterized in that The operations to query the next batch of nodes and return the generated tasks and execute include: Determine whether the next batch reaches the set threshold; If yes, report an error; if not, return the action that generates the task and executes it.

6. The method according to claim 1, characterized in that The IO type tasks are executed in parallel using a first thread pool, and the CPU type tasks are executed in parallel using a second thread pool; the number of threads in the first thread pool is greater than that in the second thread pool.

7. The method according to claim 1, characterized in that After generating and executing tasks according to each batch of nodes in the order of hierarchy from shallow to deep, it also includes: Write the execution result back to the context.

8. An electronic device, characterized in that: include: Memory and processor; Wherein, one or more computer programs are stored in the memory, and the one or more computer programs include instructions; when the instructions are executed by the processor, the electronic device executes the method according to any one of claims 1 to 7.