CN109117258A - A kind of multiple nucleus system Static task scheduling method that task based access control is mobile - Google Patents
A kind of multiple nucleus system Static task scheduling method that task based access control is mobile Download PDFInfo
- Publication number
- CN109117258A CN109117258A CN201810818312.1A CN201810818312A CN109117258A CN 109117258 A CN109117258 A CN 109117258A CN 201810818312 A CN201810818312 A CN 201810818312A CN 109117258 A CN109117258 A CN 109117258A
- Authority
- CN
- China
- Prior art keywords
- task
- dispatch list
- pending
- list
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
Abstract
The invention discloses a kind of mobile multiple nucleus system Static task scheduling methods of task based access control, this method comprises: step S10 generates the first dispatch list according to the multiple pending tasks received;Step S20 determines task moving range according to the quantity of pending task;Step S30 chooses any pending task and is used as free task point, and according to task moving range, freedom of movement task point generates the second dispatch list;Second dispatch list is denoted as the first dispatch list when determining that the second dispatch list meets task restriction condition by step S40;Step S50 sends the first dispatch list;Pending task is respectively allocated in each kernel in multiple nucleus system according to the first dispatch list and executes by step S60.Technical solution in through the invention is conducive to the dispatching effect for improving task schedule list, reduces the scheduling length of task schedule and the overall execution time during task execution, improve the operation efficiency of multi-core microprocessor.
Description
Technical field
The present invention relates to the technical fields of multi-core microprocessor task schedule, move in particular to a kind of task based access control
Dynamic multiple nucleus system Static task scheduling method.
Background technique
With the development of technology, it comes into being with multiple microprocessor, that is, multi-core processors for calculating kernel, multiple
Field is used widely, and is increasingly becoming the mainstream of computer microprocessor.Application program pair can be met using coenocytism
The requirement that multi-thread concurrent executes increases substantially microprocessor performance, but the scheduling of each task in application program inside is very big
The performance that ground affects microprocessor plays.Dispatching algorithm is meeting between task using individual task as basic execution unit
Data-driven relationship under the premise of, be that each task choosing the most suitably handles core by specific strategy, use at present
More is Static task scheduling technology, which has carried out point task by dispatching algorithm before task execution
Match, there is preferable operability, wherein list scheduling algorithm complexity in numerous schemes is minimum, and performance is significant, is always
The hot technology in multi-core microprocessor task schedule field.
And in the prior art, list scheduling algorithm is usually to carry out analysis meter to the individual task node in task list
It calculates, realization optimizes the ordering strategy of task node, but does not account for the concurrency between task node, and also no pair
Task schedule space is searched on a large scale, has certain limitation.
Summary of the invention
It is an object of the invention to: expand task schedule list search space, improve application program internal task it
Between the concurrency that executes, shorten task execution time, improve the operation efficiency of multi-core microprocessor.
The technical scheme is that a kind of multiple nucleus system Static task scheduling method that task based access control is mobile is provided,
This method comprises: step S10 generates the first dispatch list according to the multiple pending tasks received;Step S20, according to
The quantity of execution task determines task moving range;Step S30 chooses any pending task and is used as free task point, according to
Task moving range, freedom of movement task point generate the second dispatch list;Step S40 appoints when determining that the second dispatch list meets
When constraint condition of being engaged in, the second dispatch list is denoted as the first dispatch list;Step S50 sends the first dispatch list;Step S60,
Pending task is respectively allocated in each kernel in multiple nucleus system according to the first dispatch list and is executed.
In any of the above-described technical solution, further, step S10 is specifically included: step S11, more according to what is received
A pending task is generated without circuit digraph, includes multiple tasks node, any one task node pair in no circuit digraph
It should be in any one pending task;Step S12 calculates the weighted value without any of circuit digraph task node, and generates
First dispatch list.
In any of the above-described technical solution, further, step S20 is specifically included: step S21, is counted oriented without circuit
In figure in any level task node level number;The maximum value of level number is denoted as task moving range by step S22.
In any of the above-described technical solution, further, step S30 is specifically included: step S31, randomly select it is any to
Execution task is used as free task point;Step S32, the random moving direction and task moving distance for determining free task point, is moved
Dynamic direction is to be moved to the left or move right, and task moving distance is less than or equal to task moving range;Step S33, according to shifting
Dynamic direction, task moving distance, freedom of movement task point generate the second dispatch list.
In any of the above-described technical solution, further, step S40 is specifically included: step S41 judges free task point
Whether it is located at except the first dispatch list after movement, when determining that free task point is located at except the first dispatch list, executes step
Otherwise rapid S30 executes step S42;Step S42, judges whether the corresponding pending task of free task point is located at after movement
Before the corresponding previous task of the pending task, when before determining that first task is located at previous task, execution step S30,
Otherwise, step S43 is executed;Step S43, calculate the first dispatch list first task scheduling length and the second dispatch list the
Two task schedule length;Step S44, judges whether the second task schedule length is greater than or equal to first task scheduling length, when
When determining that the second task schedule length is greater than or equal to first task scheduling length, step S30 is executed, otherwise, by the second scheduling
List is denoted as the first dispatch list.
In any of the above-described technical solution, further, before step S50 further include: step S51, the second scheduling of statistics
The generation number of list;Step S52, judges whether the generation number of the second dispatch list is less than the quantity of pending task and pre-
If the product of multiple;When determining that generating number is less than product, step S30 is executed, otherwise, executes step S50.
The beneficial effects of the present invention are: generating the second dispatch list by the mobile free task point chosen, and when second
When dispatch list meets task restriction condition, the first dispatch list is replaced with the second dispatch list, is conducive to improve task schedule
The dispatching effect of list reduces the scheduling length of task schedule, by randomly selecting task node, task moving direction and appointing
Business moving distance, improves the reliability for generating the second dispatch list, dispatches by comparing first task scheduling length and second
The size of length chooses the small dispatch list of length as the dispatch list sent, is conducive to the dinger for reducing task schedule
The overall execution time during degree and task execution.
The present invention improves by calculating the weighted value without any task node in the digraph of circuit and generates the first dispatch queue
The reliability of table, by having using the maximum value of the level number of task node in no circuit digraph as task moving range
Conducive to a possibility that free task point exceeds the first dispatch list is reduced, the validity for generating the second dispatch list is improved, is mentioned
High a possibility that obtaining smaller scheduling length, improve the operation efficiency of multi-core microprocessor.
Detailed description of the invention
Above-mentioned and/or additional aspect of the invention and advantage will become from the description of the embodiment in conjunction with the following figures
Obviously and it is readily appreciated that, in which:
Fig. 1 is the mobile multiple nucleus system Static task scheduling method of task based access control according to an embodiment of the invention
Overall schematic flow diagram;
Fig. 2 is the mobile multiple nucleus system Static task scheduling method of task based access control according to an embodiment of the invention
Overall schematic flow diagram;
Fig. 3 is the schematic diagram of no circuit digraph according to an embodiment of the invention;
Fig. 4 is the mobile schematic diagram of free task point according to an embodiment of the invention.
Specific embodiment
To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific real
Applying mode, the present invention is further described in detail.It should be noted that in the absence of conflict, the implementation of the application
Feature in example and embodiment can be combined with each other.
In the following description, many details are elaborated to facilitate a thorough understanding of the present invention, still, the present invention may be used also
To be implemented using other than the one described here other modes, therefore, protection scope of the present invention is not by described below
Specific embodiment limitation.
Embodiment:
Hereinafter with reference to Fig. 1-4, embodiments of the present invention will be described.
As shown in Figure 1, the multiple nucleus system Static task scheduling method that task based access control is mobile, under overall dispatching method includes
Column step:
Step S10 generates the first dispatch list according to the multiple pending tasks received;
Step S20 determines task moving range according to the quantity of pending task;
Step S30 chooses any pending task and is used as free task point, and according to task moving range, freedom of movement is appointed
Business point, generates the second dispatch list;
Second dispatch list is denoted as the first tune when determining that the second dispatch list meets task restriction condition by step S40
Spend list;
Step S50 sends the first dispatch list;
Pending task is respectively allocated in each kernel in multiple nucleus system by step S60 according to the first dispatch list
It executes.
Fig. 2 shows the detailed implementations of above steps.
As shown in Fig. 2, the mobile multiple nucleus system Static task scheduling method detailed process packet of task based access control in the present embodiment
It includes:
Step S11 is generated without circuit digraph according to the multiple pending tasks received, is wrapped in no circuit digraph
Multiple tasks node is included, any one task node corresponds to any one pending task;
Specifically, according to the multiple pending tasks received, logical relation between parsing task constructs task mould
Type, and the task model of construction is converted into mathematical model, generate without circuit digraph (DirectedAcyclic Graph,
DAG), as shown in figure 3, including multiple tasks node P in DAG, any task node P corresponds to a pending task, task
Number in node P is corresponding node serial number, and the task node in same row constitutes a level, such as dotted line in Fig. 3
Shown in frame, the task execution rank of task node is identical in the level, and DAG can be indicated with G=(V, E, W, C), wherein V and E
The set on node and side is respectively indicated, W and C respectively indicate the set of node weights and side right weight.In the present invention, by multiple wait hold
Row task is divided into multiple levels according to the relationship between task node.
After generating DAG, the task schedule length between two neighboring task execution rank can be calculated, task schedule is long
Degree includes calculating time and the call duration time of task, as shown in figure 3, the calculating time of task 0 is 38, if task 0 and task 1
It is executed in the heart in different processor cores, then task 0 is 63 to the call duration time between task 1, corresponding, such as two are appointed
Business all executes in same processor, then call duration time is 0.Other task nodes are all and so on.
Step S12 calculates the weighted value without any of circuit digraph task node, and generates the first dispatch list.
Specifically, according to BL weight calculation formula, according to traversal principle, BL power is carried out to any task node P in DAG
Re-computation, BL weight calculation formula are as follows:
In formula, i is mission number, and w (i) refers to the calculating time of task i, c (i, j) refer to task i and task j with
The time of communication consumption is carried out between different IPs, bl (i) refers to the bl weight of task i, and meeting point refers to the task execution
It is no longer communicated with other tasks after after the completion.
Task node is subjected to ascending arrangement according to its BL weighted value, corresponding first scheduling is generated according to sequence
List, as shown in Fig. 4 (A), the first dispatch list of generation is [0,1,2,6,5,3,4,13,14,12,10,8,9,11,7].
Step S21 counts the level number without task node in any level in the digraph of circuit;
The maximum value of level number is denoted as task moving range by step S22.
Specifically, as shown in figure 3, dotted line frame indicates the task node for belonging to same level, corresponding task node
Level number is 4, in the present embodiment, and the level number of the task node of the 4th level is corresponding maximum value, maximum value 8,
Therefore, task moving range L is set as 8.
Step S31 randomly selects any pending task and is used as free task point;
Step S32, the random moving direction and task moving distance for determining free task point, moving direction are to be moved to the left
Or move right, task moving distance is less than or equal to task moving range;
Specifically, as shown in figure 4, any task node randomly selected in DAG is free task point, in the present embodiment
In, the selection corresponding task node of task 7 is free task point.Any number less than or equal to task moving range L is chosen again
Value is used as task moving distance D, and determines moving direction, in the present embodiment, the numerical value of the task moving distance D of selection at random
It is 5, the moving direction of selection is as shown in the arrow direction in Fig. 4 (B), to be moved to the left.
Step S33, according to moving direction, task moving distance, freedom of movement task point generates the second dispatch list.
Specifically, task 7 (free task point) is moved to the left 5 (moving distances), be inserted into before task 12,12 He of task
Task after task 12 translates backward, the second dispatch list is generated, as shown in Fig. 4 (C), at this point, the life of the second dispatch list
At number plus one.
More specifically, the number of iterations, initial value zero, the generation of the numerical value of the number of iterations and the second dispatch list are set
The numerical value of number is equal, after determining free task point, moving direction and task moving distance, the number of iterations is done and is added at one
Reason regenerates the second dispatch list.
Step S41 judges whether be located at except the first dispatch list after free task point is mobile,
When determining that free task point is located at except the first dispatch list, step S31 is executed, otherwise, executes step S42;
Specifically, in the present embodiment, for task point 7, if setting moving direction to move right, at this point, when moving
When dynamic task 7, task 7 will be located at except the first dispatch list, re-execute the steps S31, correspondingly, when executing step S31,
It needs again to carry out the generation number or the number of iterations of the second dispatch list plus one is handled.
Step S42, judges whether the corresponding pending task of free task point is located at the pending task pair after movement
Before the previous task answered,
When before determining that first task is located at previous task, step S31 is executed, otherwise, executes step S43;
Specifically, in the present embodiment, the previous task (the second task) of task 7 (first task) is task 3, passes through shifting
Dynamic task, generates the second dispatch list, and as shown in Fig. 4 (C), task 7 is still located at after task 3, i.e., first task is located at preposition
After task, at this point, executing step S45.
Step S43 calculates the first task scheduling length of the first dispatch list and the second task tune of the second dispatch list
Spend length;
Specifically, for calculating the first task scheduling length of the first dispatch list, by the first dispatch list to
Execution task is taken out, and according to Greedy strategy, guarantees that the task is placed on the processor core of distribution available earliest task in the heart
Time started takes all pending tasks the latest after pending task all in the first dispatch list all obtains scheduling
The task execution end time is first task scheduling length.By the corresponding first task of the first dispatch list for calculating Fig. 4 (A)
The corresponding second task schedule length of the second dispatch list that scheduling length is 417, Fig. 4 (C) is 411.
Step S44, judges whether the second task schedule length is greater than or equal to first task scheduling length;
When determining that the second task schedule length is greater than or equal to first task scheduling length, step S31 is executed, otherwise,
Second dispatch list is denoted as the first dispatch list.
Specifically, in the present embodiment, the second task schedule length is less than first task scheduling length 417 for 411, because
This, replaces the first dispatch list with the second dispatch list, at this point, the first dispatch list be [0,1,2,6,5,3,4,13,14,7,
12,10,8,9,11], the overall execution time during the scheduling length and task execution to reduce task schedule.
Step S51 counts the generation number of the second dispatch list;
Step S52, judges whether the generation number of the second dispatch list is less than the quantity and presupposition multiple of pending task
Product,
When determining that generating number is less than product, step S31 is executed, otherwise, sends the first dispatch list.
Specifically, presupposition multiple is set as 30, to guarantee the balance between performance and consumption.The second tune is generated each
When spending list, corresponding generation number or the number of iterations can all be done and add a processing, pass through the method for circulation, repeat step
S31 generates the second dispatch list, to seek the scheduling length of lesser task schedule.In the present embodiment, pending task
Quantity (task node quantity) be 15, it is therefore desirable to circulation generates, judges whether 450 the second dispatch lists meet replacement the
The requirement of one dispatch list sends the first dispatch list first task dinger corresponding with its when meeting or exceeding 450 times
Degree, at this point, the first dispatch list is the smallest second scheduling of the second task schedule length in 450 the second dispatch lists generated
List.
In the l-G simulation test of the present embodiment, using existing list scheduling algorithm compared with the method in the present invention,
By using isomorphism multiple nucleus system, it is tested for the property.The dispatch list for using existing list scheduling to generate for [0,1,2,6,
5,3,4,13,14,12,10,8,9,11,7], corresponding task schedule length is 417.And use the task schedule in the present invention
Algorithm, finally obtained dispatch list are [0,2,6,1,5,4,13,3,14,7,10,12,8,9,11], corresponding task schedule
Length is 337, and therefore, task schedule length significantly reduces, and improves the operation efficiency of multi-core microprocessor.
Further, data comparison through a large number of experiments, the task scheduling algorithm in the present invention can be improved task tune
Performance is spent, especially time ratio (average task communication time/average task execution time) is calculated in average communication and is greater than 1 feelings
Under condition, compared with prior art, scheduling performance averagely promotes 11.2% to the task scheduling algorithm in the present invention, and maximum lift is reachable
37.8%.Accordingly, with respect to existing list dispatching algorithm, method for scheduling task in the present invention is realized simple, efficiently reduces
Task schedule length, and there is higher scheduling performance, improve the operation efficiency of multi-core microprocessor.
The technical scheme of the present invention has been explained in detail above with reference to the attached drawings, mobile the invention proposes a kind of task based access control
Multiple nucleus system Static task scheduling method, comprising: according to the multiple pending tasks received, generate the first dispatch list;Root
According to the quantity of pending task, task moving range is determined;It chooses any pending task and is used as free task point, according to task
Moving range, freedom of movement task point generate the second dispatch list;When the second dispatch list of judgement meets task restriction condition
When, the second dispatch list is denoted as the first dispatch list;Send the first dispatch list;According to the first dispatch list by pending
It is engaged in executing in each kernel being respectively allocated in multiple nucleus system.Technical solution in through the invention is conducive to raising task
The dispatching effect of dispatch list reduces the scheduling length of task schedule and the overall execution time during task execution, mentions
The high operation efficiency of multi-core microprocessor.
Step in the present invention can be sequentially adjusted, combined, and deleted according to actual needs.
Unit in apparatus of the present invention can be combined, divided and deleted according to actual needs.
Although disclosing the present invention in detail with reference to attached drawing, it will be appreciated that, these descriptions are only exemplary, not
For limiting application of the invention.Protection scope of the present invention may include not departing from this hair by appended claims
For various modifications made by invention, remodeling and equivalent scheme in the case where bright protection scope and spirit.
Claims (6)
1. a kind of mobile multiple nucleus system Static task scheduling method of task based access control, which is characterized in that this method comprises:
Step S10 generates the first dispatch list according to the multiple pending tasks received;
Step S20 determines task moving range according to the quantity of the pending task;
Step S30 chooses any pending task and is used as free task point, according to the task moving range, mobile institute
Free task point is stated, the second dispatch list is generated;
Second dispatch list is denoted as institute when determining that second dispatch list meets task restriction condition by step S40
State the first dispatch list;
Step S50 sends first dispatch list;
The pending task is respectively allocated in each in multiple nucleus system by step S60 according to first dispatch list
It is executed in core.
2. the mobile multiple nucleus system Static task scheduling method of task based access control as described in claim 1, which is characterized in that described
Step S10 is specifically included:
Step S11 is generated according to the multiple pending tasks received without circuit digraph, the no circuit digraph
In include multiple tasks node, any one described task node correspond to any one described pending task;
Step S12, calculates the weighted value of task node described in any of described no circuit digraph, and generates described first and adjust
Spend list.
3. the mobile multiple nucleus system Static task scheduling method of task based access control as claimed in claim 2, which is characterized in that described
Step S20 is specifically included:
Step S21 counts the level number of task node described in any level in the no circuit digraph;
The maximum value of the level number is denoted as the task moving range by step S22.
4. the mobile multiple nucleus system Static task scheduling method of task based access control as described in claim 1, which is characterized in that described
Step S30 is specifically included:
Step S31 randomly selects any pending task as the free task point;
Step S32 determines that the moving direction and task moving distance of the free task point, the moving direction are to the left at random
Movement moves right, and the task moving distance is less than or equal to the task moving range;
Step S33, according to the moving direction, the task moving distance, the mobile free task point generates described second
Dispatch list.
5. the mobile multiple nucleus system Static task scheduling method of task based access control as described in claim 1, which is characterized in that described
Step S40 is specifically included:
Step S41 judges whether be located at except first dispatch list after the free task point is mobile,
When determining that the free task point is located at except first dispatch list, the step S30 is executed, otherwise, is executed
Step S42;
Step S42, judges whether the corresponding pending task of the free task point is located at this pending after movement
It is engaged in front of corresponding previous task,
When before determining that the first task is located at the previous task, the step S30 is executed, otherwise, executes step
S43;
Step S43, calculate first dispatch list first task scheduling length and second of second dispatch list
Business scheduling length;
Step S44, judges whether the second task schedule length is greater than or equal to the first task scheduling length,
When determining that the second task schedule length is greater than or equal to the first task scheduling length, the step is executed
Otherwise second dispatch list is denoted as first dispatch list by S30.
6. the mobile multiple nucleus system Static task scheduling method of task based access control as described in claim 1, which is characterized in that described
Before step S50 further include:
Step S51 counts the generation number of second dispatch list;
Step S52, judges whether the generation number of second dispatch list is less than the number of the pending task
The product of amount and presupposition multiple;
When determining that the generation number is less than the product, the step S30 is executed, otherwise, executes the step S50.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810818312.1A CN109117258A (en) | 2018-07-24 | 2018-07-24 | A kind of multiple nucleus system Static task scheduling method that task based access control is mobile |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810818312.1A CN109117258A (en) | 2018-07-24 | 2018-07-24 | A kind of multiple nucleus system Static task scheduling method that task based access control is mobile |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109117258A true CN109117258A (en) | 2019-01-01 |
Family
ID=64862500
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810818312.1A Pending CN109117258A (en) | 2018-07-24 | 2018-07-24 | A kind of multiple nucleus system Static task scheduling method that task based access control is mobile |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109117258A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112818003A (en) * | 2021-01-14 | 2021-05-18 | 内蒙古蒙商消费金融股份有限公司 | Execution risk estimation method and device for query task |
CN112817731A (en) * | 2021-02-25 | 2021-05-18 | 合肥工业大学 | Heterogeneous multi-core system task scheduling method based on node replication |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508708A (en) * | 2011-11-30 | 2012-06-20 | 湖南大学 | Heterogeneous multi-core energy-saving task schedule method based on improved genetic algorithm |
CN102722381A (en) * | 2011-03-25 | 2012-10-10 | 微软公司 | Techniques to optimize upgrade tasks |
CN103019835A (en) * | 2011-09-26 | 2013-04-03 | 同方股份有限公司 | System and method for optimizing interruption resources in multi-core processor |
CN103823719A (en) * | 2014-02-26 | 2014-05-28 | 杭州群核信息技术有限公司 | Distributed cloud computing system and distributed cloud computing method for executable program |
US9400682B2 (en) * | 2012-12-06 | 2016-07-26 | Hewlett Packard Enterprise Development Lp | Ranking and scheduling of monitoring tasks |
WO2018129550A1 (en) * | 2017-01-09 | 2018-07-12 | My Ally Inc. | Smart recruiting systems and associated devices and methods |
-
2018
- 2018-07-24 CN CN201810818312.1A patent/CN109117258A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102722381A (en) * | 2011-03-25 | 2012-10-10 | 微软公司 | Techniques to optimize upgrade tasks |
CN103019835A (en) * | 2011-09-26 | 2013-04-03 | 同方股份有限公司 | System and method for optimizing interruption resources in multi-core processor |
CN102508708A (en) * | 2011-11-30 | 2012-06-20 | 湖南大学 | Heterogeneous multi-core energy-saving task schedule method based on improved genetic algorithm |
US9400682B2 (en) * | 2012-12-06 | 2016-07-26 | Hewlett Packard Enterprise Development Lp | Ranking and scheduling of monitoring tasks |
CN103823719A (en) * | 2014-02-26 | 2014-05-28 | 杭州群核信息技术有限公司 | Distributed cloud computing system and distributed cloud computing method for executable program |
WO2018129550A1 (en) * | 2017-01-09 | 2018-07-12 | My Ally Inc. | Smart recruiting systems and associated devices and methods |
Non-Patent Citations (4)
Title |
---|
ARABNEJAD H: "List scheduling algorithm for heterogeneous systems by an optimistic cost table", 《IEEE》 * |
刘暾东: "面向光子网格任务调度的迭代列表算法", 《光子学报》 * |
宋宇鲲: "多核系统静态任务调度的启发式算法", 《电子测量与仪器学报》 * |
杨俊: "多核系统静态任务调度问题研究", 《中国优秀硕士论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112818003A (en) * | 2021-01-14 | 2021-05-18 | 内蒙古蒙商消费金融股份有限公司 | Execution risk estimation method and device for query task |
CN112818003B (en) * | 2021-01-14 | 2023-03-31 | 内蒙古蒙商消费金融股份有限公司 | Execution risk estimation method and device for query task |
CN112817731A (en) * | 2021-02-25 | 2021-05-18 | 合肥工业大学 | Heterogeneous multi-core system task scheduling method based on node replication |
CN112817731B (en) * | 2021-02-25 | 2022-12-06 | 合肥工业大学 | Heterogeneous multi-core system task scheduling method based on node replication |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6788019B2 (en) | Accelerate training of deep neural networks using inconsistent stochastic gradient descent | |
CN111274036B (en) | Scheduling method of deep learning task based on speed prediction | |
US8543993B2 (en) | Compiler, compile method, and processor core control method and processor | |
US9081618B2 (en) | Method and apparatus for the scheduling of computing tasks | |
CN108009016A (en) | A kind of balancing resource load control method and colony dispatching device | |
CN110389816B (en) | Method, apparatus and computer readable medium for resource scheduling | |
CN104243617B (en) | Towards the method for scheduling task and system of mixed load in a kind of isomeric group | |
CN104699461A (en) | Configuring thread scheduling on a multi-threaded data processing apparatus | |
US10956417B2 (en) | Dynamic operation scheduling for distributed data processing | |
Kamthe et al. | A stochastic approach to estimating earliest start times of nodes for scheduling DAGs on heterogeneous distributed computing systems | |
CN109117258A (en) | A kind of multiple nucleus system Static task scheduling method that task based access control is mobile | |
CN109242099A (en) | Training method, device, training equipment and the storage medium of intensified learning network | |
CN108205469A (en) | A kind of resource allocation methods and server based on MapReduce | |
CN116126346A (en) | Code compiling method and device of AI model, computer equipment and storage medium | |
Feng et al. | Robust parallel preconditioned power grid simulation on GPU with adaptive runtime performance modeling and optimization | |
JP5708003B2 (en) | Arithmetic processing device, computing node, parallel computer system, arithmetic processing method and arithmetic processing program | |
EP4280107A1 (en) | Data processing method and apparatus, device, and medium | |
CN113452546A (en) | Dynamic quality of service management for deep learning training communications | |
Miao et al. | Efficient flow-based scheduling for geo-distributed simulation tasks in collaborative edge and cloud environments | |
Yue et al. | Parameter estimation for a generalized parallel loop scheduling algorithm | |
CN110837395B (en) | Normalization processing method, device and system for multi-GPU parallel training | |
CN116932201A (en) | Multi-resource sharing scheduling method for deep learning training task | |
CN105260165A (en) | Method for dynamically controlling speculative thread number in thread-level speculation system based on SMT | |
CN113850394B (en) | Federal learning method and device, electronic equipment and storage medium | |
CN105335226B (en) | For the iterative static task list scheduling method of multicomputer system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190101 |
|
RJ01 | Rejection of invention patent application after publication |