CN108021436A

CN108021436A - A kind of process scheduling method

Info

Publication number: CN108021436A
Application number: CN201711462360.3A
Authority: CN
Inventors: 武传胜; 王丹丹
Original assignee: University of Science and Technology Liaoning USTL
Current assignee: University of Science and Technology Liaoning USTL
Priority date: 2017-12-28
Filing date: 2017-12-28
Publication date: 2018-05-11

Abstract

The present invention provides a kind of process scheduling method, and the method is the method that executive process is dispatched in a kind of synchronous driving system based on multiple sync drivers in an operating system, and the process scheduling method includes the following steps：Step 1: determined to quantify argument QA by the synchronization extent P of multiple sync drivers；Step 2: provide definite quantization argument QA to one of multiple sync drivers；The piecemeal of kernel sequence is driven to produce drive block according to argument QA is quantified.This method determines to quantify argument, wherein synchronization extent is the piecemeal quantity of the kernel sequence of corresponding multiple sync driver synchronization process to one of multiple sync drivers based on the synchronization extent in relation to multiple sync drivers；And provide and quantify argument one of to multiple sync drivers, the piecemeal of kernel sequence is driven based on quantization argument to produce drive block, this method can reach the quantization argument QA of required output process.

Description

A kind of process scheduling method

Technical field

Field of resource allocation and defined resource allocation methods technical field the present invention relates to operating system, more particularly to A kind of process scheduling method.

Background technology

For different system and aims of systems, the different dispatching method of generally use, for example, in batch processing system In, in order to look after plurality of short operation, the dispatching method of short job priority should be used；And for example in time-sharing system, in order to Guarantee system has the rational response time, should be scheduled using round robin.In a variety of dispatching methods that presently, there are, have Method be suitable for job scheduling, some methods are suitable for process scheduling；But also some dispatching methods both can be used for operation tune Degree, can be used for process scheduling.The list entries of original kernel data is received using the driver of RTOS technologies, is then exported The sequence of the kernel data driven.The scheduling argument of driver, conventionally referred to as quantifies argument (Quantitative Arguments, QA), it is the output quality after schedule driven, i.e. human eye watches the quality of output.

By RTOS, the quantization argument QA of relatively low numerical value produces the output of better quality.In addition, quantifying argument QA can also influence The size of output, the size of so-called output is the driving required bit quantity of output data.However, the size due to output data To quantify the function of the complexity (complexity) of argument QA and input data, therefore it is only indirectly to influence.Typically, The more position of more complicated kernel sequence requirements, more high quality require more positions.

Though driver is argument by quantization argument QA, actually many applications, which must not only apply, is limited in output quality On, it must also apply and be limited in the quantity of carry-out bit.If the kernel data driven is via high-speed bus N transmittable per second Transmission, driver must reach generation N per second.Since the quantity of carry-out bit can also be transfused to the complexity effect of sequence, amount The relation changed between argument QA and output size is indirect.Therefore, setting can reach the quantization argument of required output process QA is often difficult.

The content of the invention

In order to solve the problems, such as described in background technology, the present invention provides a kind of process scheduling method, and the method is one Kind be included in the method that executive process is dispatched in the synchronous driving system of multiple sync drivers, to multiple sync drivers it One, determine to quantify argument based on the synchronization extent in relation to multiple sync drivers, wherein synchronization extent is corresponding multiple synchronous drives The piecemeal quantity of the kernel sequence of dynamic device synchronization process；And provide and quantify argument one of to multiple sync drivers, based on amount Change argument to be driven the piecemeal of kernel sequence to produce drive block, this method can reach the quantization of required output process Argument QA.

In order to achieve the above object, the present invention is realized using following technical scheme：

A kind of process scheduling method, is a kind of synchronous driving system based on multiple sync drivers in an operating system The method of middle executive process scheduling, the process scheduling method include the following steps：

Step 1: determined to quantify argument QA by the synchronization extent P of multiple sync drivers；

Multiple synchronization extent P correspond to the piecemeal quantity N of the kernel sequence of multiple sync driver synchronization process, institute The piecemeal quantity N for the kernel sequence stated is the piecemeal maximum number for the kernel sequence that corresponding multiple sync drivers are capable of synchronization process Amount；

Step 2: provide definite quantization argument QA to one of multiple sync drivers；According to quantization argument QA to interior The piecemeal of core sequence is driven to produce drive block.

The step one is specially：The quantization argument QA includes：Determine to drive out by multiple sync drivers Unfinished piecemeal quantity, the foundation for beginning but not yet completing do not complete piecemeal quantity increase synchronization extent, wherein not completing block count Amount is up to the piecemeal maximum quantity of multiple sync drivers once synchronization process；

Specifically included by the synchronization extent P of the multiple sync drivers methods for determining to quantify argument QA：By synchronization extent point Gain argument is cut to determine to quantify argument, and the result of segmentation is multiplied with wrong argument；Further include：According to needed for drive block Size and actual size, update wrong argument after the piecemeal of driving kernel sequence.

The sync driver uses one of following kernel-driven operating system：VRTX, RTOS, OS/2warp and RTWINDOWS。

Compared with prior art, the beneficial effects of the invention are as follows：

The present invention a kind of process scheduling method, be it is a kind of be included in the synchronous driving system of multiple sync drivers hold The method of row process scheduling, to one of multiple sync drivers, based on the synchronization extent decision amount in relation to multiple sync drivers Change argument, wherein synchronization extent is the piecemeal quantity of the kernel sequence of corresponding multiple sync driver synchronization process；And provide Quantify argument one of to multiple sync drivers, based on quantifying argument and be driven the piecemeal of kernel sequence to drive to produce Block, this method can reach the quantization argument QA of required output process.

Brief description of the drawings

Fig. 1 is the flow chart of the process scheduling method of the present invention；

Fig. 2 is the embodiment schematic diagram of the synchronous driving system of the executive process dispatching method of the present invention；

Fig. 3 be the present invention process scheduler and sync driver processing module embodiment detailed maps；

Fig. 4 is the flow chart of the specific implementation process of the present invention；

Fig. 5 is the configuration reality that executive process is dispatched in the synchronous driving system including multiple sync drivers of the present invention Apply a system hardware structure.

Wherein：1 110- step 1 120- step 2 200- drive system 202- piecemeal 204- drive blocks of 100- processes 206- kernel sequence 210- processing module 220- process schedulers 222- quantifies argument QA 224- synchronization extent P 226- ACTUAL 228-DESIRED 240- driver 310-BEGIN_CHUNK 320-END_CHUNK 330- perform estimation block 332- performs jump module 334- modular converter 336- quantization modules 338- inverse quantification module 340- reverse conversion modules 342- datarams module 344- variable-length Drive Modules 346- exports buffer module.

Embodiment

Embodiment provided by the invention is described in detail below in conjunction with attached drawing.

It is in an operating system a kind of based on the same of multiple sync drivers as shown in Figure 1, a kind of process scheduling method The method that executive process is dispatched in drive system is walked, the process scheduling method includes the following steps：

Step 110, determined quantization argument QA by the synchronization extent P of multiple sync drivers；

Step 120, provide definite quantization argument QA to one of multiple sync drivers；According to quantization argument QA to interior The piecemeal of core sequence is driven to produce drive block.

In the description above：The quantization argument QA is a kind of scheduling argument of driver, can be used to schedule driven Export quality.Process scheduling P is the number of the quantization argument QA corresponding to the carry-out bit produced for the desired transmission process of maintenance Value.The synchronous driving system is can be using what multiple sync drivers performed multiple driving operations in same time System.

In the description above：The quantization argument QA determined by synchronization extent P can improve in synchronous driving system into The stability of journey scheduling.Especially because driver is independent operation, synchronization extent P can be used in process scheduling to avoid or reduce Quantify argument QA due to the mass efficient gain of sync driver and not with time convergent possibility.

The step 110 is specially：The quantization argument QA includes：Determine to drive out by multiple sync drivers Unfinished piecemeal quantity, the foundation for beginning but not yet completing do not complete piecemeal quantity increase synchronization extent, wherein not completing block count Amount is up to the piecemeal maximum quantity of multiple sync drivers once synchronization process；

The sync driver uses one of following kernel-driven operating system：VRTX, RTOS, OS/2 warp and RT WINDOWS。

The synchronous driving system provides high in the clouds driving service to multiple long-range users.

【Specific embodiment】

As shown in Fig. 2, it is the schematic diagram of the synchronous driving system 200 for the process scheduling method for performing Fig. 1.

Synchronous driving system 200 includes multiple processing modules 210 that can be operated at the same time.Processing module 210 includes mutual shadow Loud process scheduler 220 and driver 240, drive block 204 is driven into by the piecemeal 202 of kernel sequence 206.

Processing module 210, can be used the combination of software, hardware or both and implements.The quantity (such as N) of processing module 210 It can depend on that in application and underlying system hardware etc. be any suitable quantity.In each processing module 210, process scheduler 220 determine the numerical value of the quantization argument QA222 of driver 240, and piecemeal 202 is driven into drive block 204.

Process scheduler 1 determines the quantization argument QA of driver 1, piecemeal is based on to quantify argument QA, drives into driving Block 1 etc..

In Fig. 2, the part of Partitioning Expression of A kernel sequence, as piecemeal 202 represents the part of kernel sequence 206.Kernel sequence 206 include multiple data, may be partitioned into big block and the section including the big block of any amount etc..In fact, piecemeal 202 can be Any suitable size, as large as the size of whole kernel data, small such as big block or any Jie between.Piecemeal 202 can exist greatly Whole kernel data or small in big block.

The piecemeal 202 of different disposal module 210 shows that it can be different size with formed objects.For example, use RTOS, piecemeal 1 can be big block, and piecemeal N can be group of section or section etc..Therefore, the size of piecemeal can be considered process The module of scheduling.Piecemeal 202 can be assigned to the different processing modules 210 using any suitable distribution method.

Process scheduler 220 determines to quantify argument QA222 according to the synchronization extent P of sync driver 240.It is for example, same Step degree P224 can be determined by the quantity of 240 synchronously driven piecemeal of sync driver.Amount in synchronous driving system 220 Change the stability for the scheduling loop that argument QA222 and synchronization extent P influences the course between scheduler 220 and driver 240.

Determine that the method specific embodiment process for quantifying argument QA is as follows by the synchronization extent P of multiple sync drivers：

Exemplified by only having system, that is, N=l of a processing module 210, piecemeal 202 is successively to drive.For example, working as makes When driving i-th of piecemeal with RTOS, the numerical value that the setting of process scheduler 220 quantifies argument QA [i] is as follows：

QA [i]=GAIN*ERROR [i] (1)

In equation (1), GAIN is appropriate constant (being usually just), is performed in the case where quantifying the quality that argument is QA [i] Driver.When driver 240 produces the drive block 204 that size is ACTUAL [i] position 226, QA [i] is depended not only on, is also depended on In the complexity of kernel sequence.After i-th of piecemeal is driven, process scheduler 220 is just based on ACTUAL [i] 226 and required The size of i-th of piecemeal DESIRED [i] 228 updates wrong argument ERROR [i].

ERROR [i+l]=ERROR [i]+(ACTUAL [i]-DESIRED [i]) (2)

ERROR [i+l] to computing piecemeal i+1 QA [i+l].If driver 240 exports compare required methodically More positions (i.e. ACTUAL [i]>DESIRED [i]), ERROR [i+l] just increases with the time.Piecemeal can so be increased The numerical value of 202 quantization argument QA222, causes these piecemeals 202 using less bits to complete required output process.

To the N in Fig. 2>The synchronous driving system 200 of 1 driver, piecemeal 202 (such as piecemeal 1) are and other piecemeals 202 (such as piecemeal N) drives at the same time, and without using the feedback of other processing modules 210 (such as processing module N).The quantization argument of piecemeal 1 QA, is to be not based on feedback ACTUAL n and determine.This measure is mathematically about equivalent P times big in using compared with following step The numerical value of GAIN, wherein P are the maximum quantity for the piecemeal for being scheduled with while performing.If QA [i] in equation (1)= The calculating of GAIN*ERROR [i] is repeated in each processing module 210, and the actual gain of system 200 is about GAIN*P.Cause This, though GAIN can be the stability number of ensuing system, right synchronization system 200 is then really not so.

The importance of the selection of GAIN：Consider the example of N=1, it is assumed that the scope for quantifying argument QA is [0...10], driving Device output (10-QA) position/and per piecemeal, ERRORS is initially 0, GAIN=2 and DESIRED=5 bits/piecemeal.The display of table 1 uses ERROR [i] that equation (1) and (2) are calculated, QA [i], ACTUAL [i] i=0 ..., 5 numerical value：

Table 1：The fluctuation of quantization argument QA is caused due to instability

i	ERROR[i]	QA [i]=GAIN*ERROR [i]	ACTUAL [i]=10-QA
				0	0	0	10
1	5	10	0
				2	0	0	10
3	5	10	0
				4	0	0	10
5	5	10	0

Above-mentioned circulation is to repeat the quantization argument QA that extreme value fluctuates between 0 and 10, and is never converged to optimal (optimal) numerical value, such as QA=5.In this instance, as a result GAIN<2 numerical value finally converges to the stabilization number for quantifying argument QA Value, and in GAIN>Then from not converged when=2.Therefore, if GAIN becomes too big, between process scheduler 220 and driver 240 Schduling cycle then become unstable.

To improve the stability of the synchronous driving system 200 of Fig. 2, process scheduler 220 is according to following equation (3) with base Determine to quantify argument QA222 in synchronization extent P224：

QA=(GAIN/P) * ERROR (3)

Synchronization extent P is the maximum quantity for the piecemeal that sync driver 240 is capable of synchronization process.In the example shown in Fig. 2 In, synchronization extent P may be up to the quantity of N, i.e. driver 240.As it was previously stated, if GAIN is too big, schduling cycle becomes unstable It is fixed.No matter P why, obtain the effect of reaching stable schduling cycle by synchronization extent P to split GAIN.

In Fig. 3, include the schematic diagram of the detailed example of the processing module 210 of process scheduler 220 and driver 240. In Fig. 2 and Fig. 3, though process scheduler 220 and driver 240 are shown with the square distinguished, it is it is to be understood that process tune Spend the part of device 220 or driver 240.

Process scheduler 220 include respectively piecemeal driving perform before and after BEGIN_CHUNK () 310 and END_ CHUNK()320.Identical with Fig. 2, process scheduler 220 is to determine to quantify argument based on the synchronization extent P of driver 240 QA222 (such as QAn), drive block 204 is driven into by piecemeal 202.

Any suitable kernel-driven operating system can be used in driver 240, such as VRTX, RTOS, OS/2warp, RT WINDOWS (advanced kernel-driven (Advanced Video Coding)) etc..For example, RTOS kernel-drivens generally include Following assemblies：Perform estimation block 330, perform jump module 332, modular converter 334, quantization modules 336, inverse quantification module 338th, reverse conversion module 340, datarams module 342, variable-length Drive Module 344 and output buffer module 346.

The estimation that estimation block 330 mainly completes process amount is performed, for weighing the pretreatment of micro- process；Modular converter 334 duplicating process conversion process, are converted into the resource that processor has；The scheduling of 336 calculation procedure of quantization modules performs the time；Can Become length Drive Module 334 to convert into the long and short cycle of different processes；Export buffer module 346 reconcile different rates into The time of journey.

Kernel sequence includes a considerable amount of amount of redundancy (redundancies) between data inside, as time and space are superfluous Surplus.Data compression (intra-frame compression) technology can be reduced by conversion driving (transform coding) The spatial redundancy amount of forms data, and spatial redundancy amount can be reduced by driving the difference between prediction data and current data. In an example in figure 3, perform estimation block 330 and vector is most preferably performed to search according to past data, performing jump module 332 is To reduce time redundancy amount.Modular converter 334 performs conversion driving (such as discrete cosine transform (discrete cosine Transform)) to drive the execution jump forecasting of different pieces of information, to reduce spatial redundancy amount.

Continue modular converter 334, and the conversion of generation is that number is to quantify, the result is that by variable length by quantization modules 336 Drive Module 344 is spent to drive.Drive block 204 is stored in before transmission or storage in output buffer module 346.Contravariant vector Changing module 338 and reverse conversion module 340 is rebuild with reference to data to be used as follow-up.Perform jump module 330 Reference data can be handled so that in execution estimation block 330.

In an example in figure 3, when the numerical value for quantifying argument QA222 is given quantization by driver 240 in order to drive piecemeal 202 During module 336, driver 240 calls the numerical value that BEGIN_CHUNK () 310 quantifies argument QA222 with passback.In driving piecemeal After 202, driver 240 calls END_CHUNK (ACTUAL) 320, and wherein ACTUAL is to drive the position of piecemeal 202 actual Quantity.BEGIN_CHUNK () 310 and END_CHUNK () 320 can by the different sync drivers 240 of multiple piecemeals 202 with Any order and called, but END_CHUNK () must be after the corresponding BEGIN_CHUNK () of identical piecemeal 202 at the same time Called.

The group of any suitable software, hardware or both can be used in BEGIN_CHUNK () 310 and END_CHUNK () 320 Close and implement.For example, BEGIN_CHUNK () 310 and END_CHUNK () 320 is the function that can be performed by driver 240 Function.

Fig. 4 is the flow chart 400 according to the process scheduling method of the present invention, and the process tune of Fig. 2 and Fig. 3 has been used in method Spend device 220 and driver 240.

The flow chart 400 of the process scheduling method include by one or more flow chart elements one or more operations, Step or execution, the step 410 to 434 such as performed by driver 240, and by the step 412 that process scheduler 220 performs to 418 and step 432 to 434.In addition, different steps can be combined into less step based on required implementation, be divided into volume Outer step or elimination.Step 410 to 434 can be repeated for the piecemeal 202 that each is driven by driver 240.

In the step 410 of Fig. 4, (such as driver n) arouses (invoke) driver 240 of synchronous driving system 200 BEGIN-CHUNK (), to obtain the numerical value for quantifying argument QA (such as QAn).

In the step 412 to 418 of Fig. 4, (such as process scheduler n) performs BEGIN_CHUNK () to process scheduler 220. In step 412, counter OUTSTANDING represents the quantity for the piecemeal that driving has begun to but Yet-have is not completed, each Increase (increased) during BEGIN_CHUNK () called.In step 414, synchronization extent P initial values are 1, and foundation OUTSTANDING and increase.If there is 5 sync drivers 240, process scheduler 220 increases synchronous journey according to OUTSTANDING The numerical value of P is spent, at most increases to maximum 5.In step 416 and 418, process scheduler 220 uses equation (3) decision amount Change argument QA, and provide result to driver 240.

In the step 420 of Fig. 4, (such as driver n) is based on by 220 (such as process scheduler of process scheduler driver 240 N) the quantization argument QA (such as QAn) provided drives piecemeal 202 to produce drive block 204.Such as：, can such as the explanation with reference to Fig. 3 Using perform estimation block 330, perform jump module 332, modular converter 334, quantization modules 336, inverse quantification module 338, Reverse conversion module 340, datarams module 342, variable-length Drive Module 344 and output buffer module 346.

In the step 430 of Fig. 4, (such as driver n) arouses END_CHUMC to driver 240 after drive block 204 is produced (ACTUAL)。

In step 432 and 434, (such as process scheduler n) performs END_CHUNK (ACTUAL) to process scheduler 220. In step 432, process scheduler 220 (size (size) of ACTUALs of such as process scheduler n) based on drive block 204 and Size (size) the renewal ERROR of DESIRED.In step 434, (such as process scheduler n) is reduced process scheduler 220 (decrease) counter OUTSTANDING with point out current piecemeal driving handle completion.As pointed by 436, due to OUTSTANDING is the argument shared in multiple process schedulers 220, and BEGIN_CHUNK () will be used The newest numerical value of OUTSTANDING is when being invoked next time.

It should be noted that any suitable parallel computing can be used to the process scheduling in each processing module 210 Implement BEGIN_CHUNK () and END_CHUNK () in device 220.For example, processing module 210 can as be referred to as thread (thread) synchronization program is implemented.Furthermore since BEGIN_CHUNK () and END-CHUNK () can be by multiple drivers 240 and call at the same time, breech lock L is provided for mutual exclusion lock.In general, breech lock is procedural language structure, it allows one 240 scheduling variable of driver and prevent other drivers 240 and read or write.For example, become if a driver 240 pins OUTSTANDING and P is measured, can be updated without other drivers 240, to avoid error in data.

The synchronous driving system of the present invention can use the hardware configuration of Fig. 5, and synchronous driving system structure 500 may include to locate Manage device 510, memory 520, high speed bus interface device 540 and communication bus 530.

In Fig. 5, processor 510 is performed referring to figs. 1 to the process described in the text Fig. 4.Processor 510 can be one Processor can also be multiple.Memory 520 can be stored referring to figs. 1 to executive process in the synchronous driving system 200 described in Fig. 4 Any information needed for scheduling, the synchronization extent (P) of such as sync driver 240 and other arguments in table 2.Memory 520 The machine readable instruction fetch 524 that can be performed by processor 510 can be more stored, to make processor perform referring to figs. 1 to described in Fig. 4 Process.

Above example is implemented under premised on technical solution of the present invention, gives detailed embodiment and tool The operating process of body, but protection scope of the present invention is not limited to the above embodiments.Method therefor is such as without spy in above-described embodiment It is conventional method not mentionlet alone bright.

Claims

1. a kind of process scheduling method, is in a kind of synchronous driving system based on multiple sync drivers in an operating system The method of executive process scheduling, it is characterised in that the process scheduling method includes the following steps：

Multiple synchronization extent P correspond to the piecemeal quantity N of the kernel sequence of multiple sync driver synchronization process, described The piecemeal quantity N of kernel sequence is the piecemeal maximum quantity for the kernel sequence that corresponding multiple sync drivers are capable of synchronization process；

Step 2: provide definite quantization argument QA to one of multiple sync drivers；According to quantization argument QA to kernel sequence The piecemeal of row is driven to produce drive block.

2. a kind of process scheduling method according to claim 1, it is characterised in that the step one is specially：It is described Quantization argument QA include：The unfinished piecemeal quantity completed determining to have driven by multiple sync drivers but not yet, according to Increase synchronization extent according to unfinished piecemeal quantity, once synchronously located wherein not completing piecemeal quantity and being up to multiple sync drivers The piecemeal maximum quantity of reason；

Specifically included by the synchronization extent P of the multiple sync drivers methods for determining to quantify argument QA：Split by synchronization extent and increased The result of segmentation is multiplied by beneficial argument with determining to quantify argument with wrong argument；Further include：According to the required size of drive block And actual size, wrong argument is updated after the piecemeal of driving kernel sequence.

3. a kind of process scheduling method according to claim 1, it is characterised in that the sync driver uses following One of kernel-driven operating system：VRTX, RTOS, OS/2warp and RT WINDOWS.