CN104899089A - Task scheduling method in heterogeneous multi-core architecture - Google Patents
Task scheduling method in heterogeneous multi-core architecture Download PDFInfo
- Publication number
- CN104899089A CN104899089A CN201510270127.XA CN201510270127A CN104899089A CN 104899089 A CN104899089 A CN 104899089A CN 201510270127 A CN201510270127 A CN 201510270127A CN 104899089 A CN104899089 A CN 104899089A
- Authority
- CN
- China
- Prior art keywords
- task
- task scheduling
- core
- scheduling
- data
- 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
Abstract
The present invention relates to a task scheduling method in a heterogeneous multi-core architecture. Between a plurality of cores, data transmission is implemented in a manner of sharing a memory, and information transmission is implemented in a manner of interruption and polling; the shared memory is partitioned into a command area, a result area, a data input area and a data output area, and high-efficiency data structures are designed according to functions of these areas so as to establish a heterogeneous multi-core interactive memory model; and task scheduling adopts a plurality of circular queues to implement priority task queues, so that a high-efficiency fair task scheduling mechanism is implemented. The present invention provides a task scheduling method, which has high efficiency, good expandability and high portability, for the master-slave heterogeneous multi-core architecture. With the task scheduling method, load of a main core can be greatly reduced and integral load on a system chip can be balanced. Moreover, the task scheduling method is low in call overhead, which can not only efficiently meet the scheduling requirement of a single task, but also simultaneously meet the real-time scheduling requirement of a plurality of tasks.
Description
Technical field
The present invention relates to field of computer architecture, particularly relate to a kind of method for scheduling task towards master-slave mode heterogeneous multi-core system.
Background technology
As time goes on, people to the requirement of calculated performance and efficiency in continuous lifting.The lifting of early stage single core processor performance, has benefited from semiconductor technology and the development at full speed of large scale integrated circuit technology.And now, be limited by the constraint of power consumption and design complexities, the demand that the performance boost of single core processor has been out of step with the times.
The performance bottleneck breaking through single core processor that appears as of isomorphism polycaryon processor provides good solution.Isomorphism polycaryon processor has become the main flow of CPU design and Manufacturing Technology Development now, no matter is at mobile computing platform, desktop computing platform, or large scale computer computing platform, and isomorphism polycaryon processor all plays its vital effect.
Meanwhile, people also rise to a new height to processor in the demand of special field calculated performance, the such as broadcasting of ultra high-definition video, the real-time rendering of big game, parallel computation of large data etc.And application scenarios difference is good at by the processor of different architecture, these are towards the demand of specific application area, have promoted the development of heterogeneous multi-nucleus processor.In addition, between processor, isomery characteristic can realize the instruction isolation between different architecture, also contributes to the safety problem solving computer realm.
Process due to different architecture is internuclear to differ greatly, scheduling of how finishing the work efficiently, reduces call overhead, brings great challenge all to the interaction mechanism of heterogeneous multi-core system, internal memory division etc. designs.
According to the status of each process core, current heterogeneous multi-core system is mainly divided into the heterogeneous polynuclear of the heterogeneous polynuclear of master-slave mode and non-master-slave mode.In master-slave mode heterogeneous multi-core system, operating system only operates on main core, and other process cores, as auxiliary kernel, assist main core to finish the work.In non-master-slave mode heterogeneous multi-core system, the status of each process core is identical, and operating system may operate on any one process core.
The realization of non-master-slave mode heterogeneous multi-core system is complicated, and technology maturity is not high, and one of them example is the HSA framework (Heterogeneous System Architecture) promoted by AMD, is not also widely used at present.By contrast, master-slave mode heterogeneous multi-core system then has a lot of successfully example, and as the heterogeneous polynuclear structure of CPU+DSPs and CPU+GPU, their portability is strong, and has a lot of ripe programming model (OpenCL, CUDA etc.).
Summary of the invention
The technical problem to be solved in the present invention is: propose a kind of efficient method for scheduling task under the heterogeneous multi-core system of master slave mode.
The technical solution adopted in the present invention is: a kind of method for scheduling task towards heterogeneous multi-core system, realizes the transmission of data by the mode of shared drive between multinuclear, by interrupt and the mode of poll realizes the transmission of information; Shared drive is divided into command area, fruiting area, Data input section and data output area, and according to the efficient data structure of the Functional Design of regional, sets up memory model mutual between heterogeneous polynuclear; Task scheduling adopts multiple circle queue to realize priority tasks queue, realizes efficient, fair Task Scheduling Mechanism;
Concrete steps comprise:
1) guide unloading phase, internuclear interruption as the interactive mode between multinuclear, make multinuclear separately internal memory initialization, boot from copy and amendment total line by initialization operation synchronous;
2) at task scheduling phase, between main core and auxiliary kernel, data transmission is realized by the mode of shared drive; Main core and auxiliary kernel on this basis, realize the transmission of mission bit stream by the mode of poll.
2. a kind of method for scheduling task towards heterogeneous multi-core system according to claim 1, it is characterized in that: the driving of described task scheduling, on the one hand for application program provides abstract calling interface, receives and manages the heterogeneous task computation requests of application program submission; On the other hand, task scheduling drive carries out alternately, completing the transmission of information and data between main core and auxiliary kernel with the mode of poll and shared drive, with efficient, fair Task Scheduling Mechanism by task scheduling to auxiliary kernel.
The present invention has the heterogeneous multi-core system of master-slave mode: main core operation system and task scheduling drive, task scheduling drives is responsible for the scheduling of task and the reception of result, multiple auxiliary kernel function monitor program, monitor is responsible for assisting main core to perform being scheduled of task.
The invention has the beneficial effects as follows: for master-slave mode heterogeneous multi-core system provide a kind of efficient, extensibility is good, portable strong method for scheduling task; Adopt the present invention significantly can reduce main core load, the overall load on equalizing system chip; And this method for scheduling task call overhead is little, not only can meet the dispatching requirement of single task efficiently, the Real-Time Scheduling demand of multitask can also be met simultaneously.
Accompanying drawing explanation
Below in conjunction with drawings and Examples, the present invention is further described.
Fig. 1 is that the multinuclear under heterogeneous multi-core system of the present invention guides Booting sequence figure;
Fig. 2 is the task interactive frame figure under heterogeneous multi-core system of the present invention;
Fig. 3 is the interaction diagrams under heterogeneous multi-core system of the present invention;
Fig. 4 is the frame diagram of the Task Scheduling Mechanism under heterogeneous multi-core system of the present invention;
Fig. 5 is the design drawing of the task scheduling under heterogeneous multi-core system of the present invention.
Embodiment
The present invention is further detailed explanation with preferred embodiment by reference to the accompanying drawings now.These accompanying drawings are the schematic diagram of simplification, only basic structure of the present invention are described in a schematic way, and therefore it only shows the formation relevant with the present invention.
Fig. 1 illustrates heterogeneous polynuclear and guides the process flow diagram started: after System on Chip/SoC powers on, and each processor core can jump to and guide enabling address to start fetching execution.
Owing to needing to carry out, from copy, can accessing operation be produced, and this operation after main core completes internal memory initialization, just need to be carried out in the guiding start-up course of auxiliary kernel.Therefore, respective guiding Booting sequence can not be completed independently between main core and auxiliary kernel, need to set up a set of interaction mechanism and ensure that following operation is carried out smoothly: (1) auxiliary kernel, after main core completes internal memory initialization, just can carry out accessing operation.(2) main core need wait for auxiliary kernel complete Bootstrap Commissioning Program from copy after, could revise total line by.
In order to ensure above synchronous operation, the present invention by internuclear interruption GPIO with meet heterogeneous polynuclear guide unloading phase interaction demand.The register of GPIO, after completing internal memory initialization operation, is set to height by main core.Auxiliary kernel then by check this register whether set high, judge whether internal memory initialization completes, guide start-up routine from copy function.After GPIO register is set to height by main core, its Bootstrap Commissioning Program also can complete from copying accessing operation.Complete after copy at main core, configuration bus route need be revised.In order to avoid total line is by after revising, auxiliary kernel cannot access Bootstrap Commissioning Program, so main core need wait for that auxiliary kernel completes after copy, could revise configuration bus route.Here be also ensure that this is synchronous by internuclear interruption GPIO.The register of GPIO, completing after copy, is set to low by auxiliary kernel.Main core by check this register whether set low, judge whether can configuration modification total line by.Afterwards, heterogeneous polynuclear can independently complete initialization procedure follow-up separately.
After completing heterogeneous polynuclear initialization startup, between main core and auxiliary kernel, completed the transmission of data by the mode of shared drive.Shared drive region is divided into Bootstrap Commissioning Program district, monitor district, command area, fruiting area, Data input section, data output area and reserved area composition: (1) Bootstrap Commissioning Program district: for depositing the Bootstrap Commissioning Program of auxiliary kernel from copy; (2) monitor district: for depositing the supervisory programme operated on auxiliary kernel, it is responsible for carrying out data interaction with main core, and the tasks carrying on management auxiliary kernel; (3) command area: for depositing the task order that main core sends to auxiliary kernel, it adopts the data structure of circle queue to safeguard; (4) fruiting area: for depositing the result phase that auxiliary kernel returns to main core, it adopts the data structure of circle queue to safeguard; (5) Data input section: for depositing in the relevant data of task; (6) data output area: for depositing in the relevant data of result; (7) reserved area: as reservation, for subsequent expansion.When main core sends task to auxiliary kernel, auxiliary kernel needs make corresponding in time and read and write order and result, and therefore command area and fruiting area are all the states being configured to Non-Cacheable.And the high efficiency of other regions in order to read and write, be all configured to the state of Cacheable.
The interactive frame that Fig. 2 illustrates heterogeneous polynuclear carries out the mutual of task, and in this framework, auxiliary kernel, as the auxiliary computing device of main core, receives being sended over by shared drive by main core of task.From the angle of main core, logic that whole interactive frame meets " reception of task transmission/result ": main core sends the task of needs execution to auxiliary kernel, receives the result of tasks carrying generation simultaneously from auxiliary kernel.On the memory model introduced before whole interactive frame is based upon.The level calling end at main core and auxiliary kernel is separately abstract as follows: (1) main core end layer is secondary abstract: the level that calls of main core is from top to bottom.The part needing auxiliary kernel in application program to have assisted, produces corresponding system call by heterogeneous task storehouse.System call be absorbed in kernel state task scheduling drive, the corresponding task demand of moving is converted to a series of task transmit operation, by task dispatcher in the mode of poll by job invocation to the corresponding region in shared drive.And the part receiving the complete result of auxiliary kernel is needed in main core application program, produce corresponding system call by heterogeneous task storehouse equally.The task scheduling that system call is absorbed in kernel state drives, and by task dispatcher in the mode of poll, is read in the corresponding region of corresponding execution result from shared drive, finally, returns to application program through a series of encapsulation.(2) auxiliary kernel end layer is secondary abstract: the level that calls of auxiliary kernel is from the bottom to top.Supervisory programme is read by the mode of poll and resolves the task in shared drive, obtains the data that task is corresponding.Then by searching task list, the respective function of task being given in function library performs, and execution result and state is returned to supervisory programme.Result and state, after checking execution result and state and encapsulate, are put to corresponding shared drive region by supervisory programme.
Fig. 3 illustrates the interaction flow of heterogeneous polynuclear: auxiliary kernel enters poll waiting status after completing and starting initialization, by poll specified memory zone bit, judges whether to need to load monitor program.Now main core does not affect by auxiliary kernel, normally works alone, operation system.When main core needs auxiliary kernel to assist Heterogeneous Computing task, by driving monitor binary program cross compile on main core completed, being copied to and specifying shared drive monitor district, and revising corresponding zone bit.Auxiliary kernel, after detecting that zone bit is modified, just jumps to and specifies shared drive monitor district, instruction fetch, function monitor program.Monitor program, by poll check command area, has judged whether that task arrives.Time main core is revised in complement mark position, check that whether return state is correct, if correct, prove that monitor program has loaded, task scheduling can be carried out by driving.The monitor program that auxiliary kernel runs, once detect that task arrives behind command area, just reads this task, then gives respective function in function library to perform this task, finally execution result is returned to driving.The execution result received is returned to application program by the driver on main core, thus completes heterogeneous polynuclear task scheduling flow process.
Fig. 4 illustrates the frame diagram of heterogeneous polynuclear Task Scheduling Mechanism, and task dispatcher is carried out alternately with application program and auxiliary kernel respectively by system call and mutual two aspects of shared drive, thus completes the task scheduling of heterogeneous polynuclear.No matter be in system call aspect, or in the mutual aspect of shared drive, main logic to be processed needed for task dispatcher is all task " reception of task transmission/result " logic: (1) system call: abstract as system layer of task dispatcher, need for application providing system calling interface, and call while supporting multiple program.From the angle of application program, it is by simple abstraction interface, can the transmission of finishing the work of convenience and high-efficiency and the reception operation of result.From the angle of task dispatcher, it has needed the inside of abstraction interface to realize.When task sends, the inside of abstraction interface realizes the encapsulation, the encapsulation of task and the maintenance of task queue that comprise system call.This part includes task definition by the memory copying of User space to kernel state, the operations such as the internal memory initialization of task queue and the maintenance of enqueue operations atomicity.When reception result, the inside of abstraction interface realizes the encapsulation, the encapsulation of internal memory result and the maintenance of process queue that comprise system call.This part includes the sleep of consumer process and wakes control up, and result is mated with consumer process, and resultant content is by the operation such as memory copying of kernel state to User space.(2) shared drive is mutual: task dispatcher sends task and abstract from the centre of auxiliary kernel reception result as to auxiliary kernel, has needed a large amount of shared drive interactive operations, thus it is mutual to complete with auxiliary kernel.When the task of transmission, task dispatcher needs to finish the work the scheduling of task in queue, and task is encapsulated, and is copied to command area and the Data input section of shared drive.This part includes out the maintenance of team's operation atomicity, and task is by the operation such as copy of kernel state internal memory to I/O internal memory.When result receives, task dispatcher has needed the fruiting area of result from shared drive and the extraction of data output area, the maintenance of process queue.This part includes result has I/O internal memory to the copy of kernel state internal memory, the operations such as control that wake up and sleep of process queue.
Fig. 5 is the design drawing of the task scheduling illustrated under heterogeneous multi-core system, what the present invention adopted is the scheduling strategy of priority scheduling, utilize the structure of multitask queue to represent multipriority, each queue corresponds to a priority, its specific design is as follows: multitask queue has multiple circle queue safeguarded by task dispatcher to form, and each task queue is corresponding and a task priority.Task can belong to i+1 the priority of one of 0 ~ i, and each priority j corresponds to a task queue queue_j.And each task queue can deposit k task at most simultaneously.If task queue is full, newly entering of task then can be joined the team unsuccessfully: if the process of unblock formula access, can return corresponding status code immediately; If the process of block type access, then after waiting list non-NULL, people's task can be put according to FIFO order.
On the basis of the structure of a upper joint multitask queue, detailed design four key operations: the selection of (1) task queue: task completes encapsulation when entering task dispatcher, directly selects corresponding task queue according to its priority.The time complexity of this operation is O (1).(2) enqueue operations of task: check circle queue state, then direct region of memory task being put into circle queue write pointers point, and upgrade write pointer.The time complexity of this operation is O (1).(3) selection of next task: travel through the queue to lowest priority successively from the queue of limit priority, until the queue of the non-NULL found, this queue is the selection queue of next task.The time complexity of this operation is O (1).(4) taking-up of next task: from the circle queue that last action is chosen, takes out the task that current read pointer is pointed to, and upgrades read pointer.The time complexity of this operation is O (1).
The just the specific embodiment of the present invention described in above instructions, various illustrating is not construed as limiting flesh and blood of the present invention, person of an ordinary skill in the technical field after having read instructions can to before described embodiment make an amendment or be out of shape, and do not deviate from the spirit and scope of the invention.
Claims (2)
1. towards a method for scheduling task for heterogeneous multi-core system, it is characterized in that: the transmission being realized data between multinuclear by the mode of shared drive, by interrupt and the mode of poll realizes the transmission of information; Shared drive is divided into command area, fruiting area, Data input section and data output area, and according to the efficient data structure of the Functional Design of regional, sets up memory model mutual between heterogeneous polynuclear; Task scheduling adopts multiple circle queue to realize priority tasks queue, realizes efficient, fair Task Scheduling Mechanism;
Concrete steps comprise:
1) guide unloading phase, internuclear interruption as the interactive mode between multinuclear, make multinuclear separately internal memory initialization, boot from copy and amendment total line by initialization operation synchronous;
2) at task scheduling phase, between main core and auxiliary kernel, data transmission is realized by the mode of shared drive; Main core and auxiliary kernel on this basis, realize the transmission of mission bit stream by the mode of poll.
2. a kind of method for scheduling task towards heterogeneous multi-core system according to claim 1, it is characterized in that: the driving of described task scheduling, on the one hand for application program provides abstract calling interface, receives and manages the heterogeneous task computation requests of application program submission; On the other hand, task scheduling drive carries out alternately, completing the transmission of information and data between main core and auxiliary kernel with the mode of poll and shared drive, with efficient, fair Task Scheduling Mechanism by task scheduling to auxiliary kernel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510270127.XA CN104899089A (en) | 2015-05-25 | 2015-05-25 | Task scheduling method in heterogeneous multi-core architecture |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510270127.XA CN104899089A (en) | 2015-05-25 | 2015-05-25 | Task scheduling method in heterogeneous multi-core architecture |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104899089A true CN104899089A (en) | 2015-09-09 |
Family
ID=54031767
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510270127.XA Pending CN104899089A (en) | 2015-05-25 | 2015-05-25 | Task scheduling method in heterogeneous multi-core architecture |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104899089A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107391245A (en) * | 2017-07-18 | 2017-11-24 | 致象尔微电子科技（上海）有限公司 | A kind of software systems of multi core chip |
| CN107491355A (en) * | 2017-08-17 | 2017-12-19 | 山东浪潮商用系统有限公司 | Funcall method and device between a kind of process based on shared drive |
| CN108319510A (en) * | 2017-12-28 | 2018-07-24 | 大唐软件技术股份有限公司 | A kind of isomery processing method and processing device |
| CN108984280A (en) * | 2017-06-05 | 2018-12-11 | 深圳市中兴微电子技术有限公司 | A kind of management method and device, computer readable storage medium of chip external memory |
| CN109426572A (en) * | 2017-08-29 | 2019-03-05 | 杭州海康威视数字技术股份有限公司 | Task processing method, device and electronic equipment |
| CN109542832A (en) * | 2018-11-30 | 2019-03-29 | 青岛方寸微电子科技有限公司 | Communication system and method between a kind of heterogeneous polynuclear CPU of no lock mechanism |
| CN109558250A (en) * | 2018-11-02 | 2019-04-02 | 锐捷网络股份有限公司 | A kind of communication means based on FPGA, equipment, host and isomery acceleration system |
| CN110262900A (en) * | 2019-06-20 | 2019-09-20 | 山东省计算中心（国家超级计算济南中心） | Lock synchronous operation accelerated method is communicated between a kind of main core based on Shen prestige many-core processor and core group |
| CN110990151A (en) * | 2019-11-24 | 2020-04-10 | 浪潮电子信息产业股份有限公司 | Service processing method based on heterogeneous computing platform |
| CN111176828A (en) * | 2018-11-09 | 2020-05-19 | 三星电子株式会社 | System on chip comprising multi-core processor and task scheduling method thereof |
| CN111338649A (en) * | 2020-02-14 | 2020-06-26 | 浪潮商用机器有限公司 | Heterogeneous system, acceleration method and device thereof and readable storage medium |
| CN111459647A (en) * | 2020-06-17 | 2020-07-28 | 北京机电工程研究所 | DSP multi-core processor parallel operation method and device based on embedded operating system |
| CN112068955A (en) * | 2020-08-21 | 2020-12-11 | 北京科技大学 | Communication optimization method in heterogeneous multi-core platform processor and electronic equipment |
| WO2021022964A1 (en) * | 2019-08-02 | 2021-02-11 | 中兴通讯股份有限公司 | Task processing method, device, and computer-readable storage medium based on multi-core system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070288739A1 (en) * | 2006-06-09 | 2007-12-13 | Dale Jason N | System and method for masking a boot sequence by running different code on each processor |
| CN101354693A (en) * | 2008-09-11 | 2009-01-28 | 重庆邮电大学 | Communication scheduling system and method among cores of isomerization multi-core processor |
| CN101464807A (en) * | 2009-01-08 | 2009-06-24 | 杭州华三通信技术有限公司 | Application program loading method and device |
| CN101533355A (en) * | 2009-04-15 | 2009-09-16 | 杭州华三通信技术有限公司 | Method for starting multioperation system and device thereof |
| CN102713852A (en) * | 2012-02-01 | 2012-10-03 | 华为技术有限公司 | Multicore processor system |
| CN103810139A (en) * | 2014-01-24 | 2014-05-21 | 浙江众合机电股份有限公司 | Data exchange method and device for multiple processors |
| CN104461716A (en) * | 2014-12-29 | 2015-03-25 | 迈普通信技术股份有限公司 | Access method of multi-nucleus heterogeneous system and multi-nucleus heterogeneous system |
-
2015
- 2015-05-25 CN CN201510270127.XA patent/CN104899089A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070288739A1 (en) * | 2006-06-09 | 2007-12-13 | Dale Jason N | System and method for masking a boot sequence by running different code on each processor |
| CN101354693A (en) * | 2008-09-11 | 2009-01-28 | 重庆邮电大学 | Communication scheduling system and method among cores of isomerization multi-core processor |
| CN101464807A (en) * | 2009-01-08 | 2009-06-24 | 杭州华三通信技术有限公司 | Application program loading method and device |
| CN101533355A (en) * | 2009-04-15 | 2009-09-16 | 杭州华三通信技术有限公司 | Method for starting multioperation system and device thereof |
| CN102713852A (en) * | 2012-02-01 | 2012-10-03 | 华为技术有限公司 | Multicore processor system |
| CN103810139A (en) * | 2014-01-24 | 2014-05-21 | 浙江众合机电股份有限公司 | Data exchange method and device for multiple processors |
| CN104461716A (en) * | 2014-12-29 | 2015-03-25 | 迈普通信技术股份有限公司 | Access method of multi-nucleus heterogeneous system and multi-nucleus heterogeneous system |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108984280A (en) * | 2017-06-05 | 2018-12-11 | 深圳市中兴微电子技术有限公司 | A kind of management method and device, computer readable storage medium of chip external memory |
| CN107391245A (en) * | 2017-07-18 | 2017-11-24 | 致象尔微电子科技（上海）有限公司 | A kind of software systems of multi core chip |
| CN107491355A (en) * | 2017-08-17 | 2017-12-19 | 山东浪潮商用系统有限公司 | Funcall method and device between a kind of process based on shared drive |
| CN109426572A (en) * | 2017-08-29 | 2019-03-05 | 杭州海康威视数字技术股份有限公司 | Task processing method, device and electronic equipment |
| CN109426572B (en) * | 2017-08-29 | 2021-07-02 | 杭州海康威视数字技术股份有限公司 | Task processing method and device and electronic equipment |
| CN108319510A (en) * | 2017-12-28 | 2018-07-24 | 大唐软件技术股份有限公司 | A kind of isomery processing method and processing device |
| CN109558250A (en) * | 2018-11-02 | 2019-04-02 | 锐捷网络股份有限公司 | A kind of communication means based on FPGA, equipment, host and isomery acceleration system |
| CN111176828A (en) * | 2018-11-09 | 2020-05-19 | 三星电子株式会社 | System on chip comprising multi-core processor and task scheduling method thereof |
| CN111176828B (en) * | 2018-11-09 | 2024-04-05 | 三星电子株式会社 | System on chip comprising multi-core processor and task scheduling method thereof |
| CN109542832A (en) * | 2018-11-30 | 2019-03-29 | 青岛方寸微电子科技有限公司 | Communication system and method between a kind of heterogeneous polynuclear CPU of no lock mechanism |
| CN110262900A (en) * | 2019-06-20 | 2019-09-20 | 山东省计算中心（国家超级计算济南中心） | Lock synchronous operation accelerated method is communicated between a kind of main core based on Shen prestige many-core processor and core group |
| CN110262900B (en) * | 2019-06-20 | 2023-09-29 | 山东省计算中心（国家超级计算济南中心） | Synchronous operation acceleration method for communication lock between main core and core group based on Shenwei many-core processor |
| WO2021022964A1 (en) * | 2019-08-02 | 2021-02-11 | 中兴通讯股份有限公司 | Task processing method, device, and computer-readable storage medium based on multi-core system |
| CN110990151A (en) * | 2019-11-24 | 2020-04-10 | 浪潮电子信息产业股份有限公司 | Service processing method based on heterogeneous computing platform |
| CN111338649A (en) * | 2020-02-14 | 2020-06-26 | 浪潮商用机器有限公司 | Heterogeneous system, acceleration method and device thereof and readable storage medium |
| CN111459647A (en) * | 2020-06-17 | 2020-07-28 | 北京机电工程研究所 | DSP multi-core processor parallel operation method and device based on embedded operating system |
| CN112068955A (en) * | 2020-08-21 | 2020-12-11 | 北京科技大学 | Communication optimization method in heterogeneous multi-core platform processor and electronic equipment |
| CN112068955B (en) * | 2020-08-21 | 2023-10-27 | 北京科技大学 | Communication optimization method in heterogeneous multi-core platform processor and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104899089A (en) | Task scheduling method in heterogeneous multi-core architecture | |
| CN102508712B (en) | Middleware system of heterogeneous multi-core reconfigurable hybrid system and task execution method thereof | |
| JP6525286B2 (en) | Processor core and processor system | |
| CN102360309B (en) | Scheduling system and scheduling execution method of multi-core heterogeneous system on chip | |
| CN102129390B (en) | Task scheduling system of on-chip multi-core computing platform and method for task parallelization | |
| CN105242962B (en) | The quick triggering method of lightweight thread based on isomery many-core | |
| US9582320B2 (en) | Computer systems and methods with resource transfer hint instruction | |
| CN103279445A (en) | Computing method and super-computing system for computing task | |
| KR20150122178A (en) | Executing an operating system on processors having different instruction set architectures | |
| CN112463709A (en) | Configurable heterogeneous artificial intelligence processor | |
| CN102087609A (en) | Dynamic binary translation method under multi-processor platform | |
| Charitopoulos et al. | Hardware task scheduling for partially reconfigurable FPGAs | |
| US7565659B2 (en) | Light weight context switching | |
| JP2008186175A (en) | Start control method of operating system and information processor | |
| CN107180010A (en) | Heterogeneous computing system and method | |
| CN111796921A (en) | Embedded multi-core operating system scheduling method, scheduling device, electronic equipment and storage medium | |
| US20150007170A1 (en) | Systems and Methods for Providing Hypercall Interface for Virtual Machines | |
| CN104346132A (en) | Control device applied to running of intelligent card virtual machine and intelligent card virtual machine | |
| CN101976204A (en) | Service-oriented heterogeneous multi-core computing platform and task scheduling method used by same | |
| CN103262039A (en) | Methods and systems for synchronous operation of a processing device | |
| WO2016028425A1 (en) | Programmatic decoupling of task execution from task finish in parallel programs | |
| WO2021143305A1 (en) | Software module integration system for numerical application and operation method | |
| WO2021098257A1 (en) | Service processing method based on heterogeneous computing platform | |
| CN101996105A (en) | Static software/hardware task dividing and dispatching method for reconfigurable computing platform | |
| CN101349975B (en) | Method for implementing interrupt bottom semi-section mechanism in embedded operation system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150909 |