CN85104907B - Hierarchy multprocessor system and method of control - Google Patents
Hierarchy multprocessor system and method of control Download PDFInfo
- Publication number
- CN85104907B CN85104907B CN85104907A CN85104907A CN85104907B CN 85104907 B CN85104907 B CN 85104907B CN 85104907 A CN85104907 A CN 85104907A CN 85104907 A CN85104907 A CN 85104907A CN 85104907 B CN85104907 B CN 85104907B
- Authority
- CN
- China
- Prior art keywords
- regional area
- local management
- management program
- processing
- program
- 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.)
- Expired
Links
Images
Abstract
The present invention relates to a multiprocessor system in which a plurality of processors share a main memory. The multiprocessor system is provided with two operation systems, wherein one operation system is a local management program executes processing in a regional area; the other operation system is a global management program which needs to execute access so as to execute processing. When the local management program executes processing and needs to access an area beyond the regional area, the local management program can make a request to the global management program to execute processing. In this way, the local management program only executes processing in the regional area; more than one processor can not simultaneously execute processing in the regional area.
Description
The relevant a kind of multicomputer system of the present invention by the shared primary memory of some processors (CPU).
As everyone knows, a computer system is provided with operating system (OS), system resource is effectively used, and computer system is turned round effectively.
In the multicomputer system of prior art, the control that the shared primary memory of a plurality of CPU, OS must possess when providing the execution operation to CPU.In this case, will run into following problem:
(1) in the zone at primary memory during access, a blockade control of forbidding other CPU access on primary memory need be arranged in order to carry out an operation as a CPU, this operation is complicated, and the maintenance of OS is also pretty troublesome.
(2) block often locking and release.So just increased the overhead of OS.
(3) because lock contention is arranged,, can not improve performance even increase the quantity of CPU.
Purpose of the present invention provides a kind of multicomputer system and control method thereof, even so that all increase at the number of CPU, the design of its OS and safeguard all very easily and minimizes the CPU competition.
In the present invention, OS is divided into two levels, i.e. local management routine processes local resource, global administration's routine processes global resource.These two supervisory routines are carried out by two CPU respectively.The CPU that is assigned to the local management program is called local cpu, and the CPU that is assigned to global administration's program is called overall CPU.Because the local management program does not need to consider the existence of other CPU, therefore just do not need blockade is locked and discharges, so the design of OS can be the same with the OS of no multiprocessor function easy with maintenance.Owing to reduced dependence, thereby reduced competition (can not be subjected to the interference of any other CPU when user program and local management procedure operation) with other CPU, just proportional raising of the performance of system like this with the number increase of CPU to supervisory routine.
The accompanying drawing summary:
Fig. 1 represents the general construction form of one embodiment of the invention,
Fig. 2 is expressed as the control flow chart of a local management program,
Fig. 3 represents a kind of form of a GETMAIN instruction,
Fig. 4 represents a control block,
The process flow diagram that Fig. 5 controls for explanation global administration program,
Fig. 6,7 and 8 show some examples of firmware control.
Fig. 1 represents the general structure of one embodiment of the invention.CPU2-1,2-2,3-1, the shared primary memory 1 of 3-2 and 3-3.Primary memory 1 has regional area 12-1 ... 12-N, wherein only use when the CPU of appointment handles in each zone, and other has a common area 11, is used jointly when handling by regional area.Include a local management program 16 in each regional area 12.Regional area 12 comprises an operation exclusive district 17 again, and it comprises a local resource 19 and an application program 18.When by local management program 16 and supervisory routine 13 control executive utilities, just generation-operation in computer system.Each local cpu is only handled in a regional area, and when handling in a regional area, can not be carried out simultaneously by some CPU2 and 3.For example, local cpu 3-1 carries out the processing of regional area 12-1, and local cpu 3-2 carries out the processing of regional area 12-2, and local cpu 3-3 carries out the processing in the regional area 12-3.Other regional area is not handled at this moment.When only allowing other regional area to handle after a local cpu 3 has been finished processing of corresponding topical zone or Interrupt Process, just handles other regional area.Local cpu 3 does not rely on other CPU when handling and carries out separately.Local management program 16 is very simple, as a system that has only a CPU, and can make full use of the operation of CPU.
For executive utility 18, not only can in a regional area, handle, and can utilize for the common global resource that uses of all regional areas and handle.In the time of need carrying out this processing, local management program 16 request global administrations programs 13 are carried out.Global administration's program 13 is carried out by two overall CPU 2-1 and 2-2.Global administration's program can or be carried out by the CPU of any number by one.The result of global administration's program 13 is sent back to the regional area of the local management of filing a request.When global administration's program 13 is being handled,, forbid the relevant processing of being carried out in regional area and global administration program 13 of processing of local management program 16 for the competition between the processing of the processing that prevents global administration's program and local management program.
In the present embodiment, the structure of overall CPU 2 and local cpu 3 is identical, just the zones of different of their accessing main memory.
Referring to Fig. 2 to Fig. 8 embodiment of the present invention are done to explain in detail below.
We suppose that local cpu 3-1 during executive utility 18, sends a request distributes a workspace on primary memory instruction GETMAIN in regional area 12-1.At this moment just send one to local management program 16-1 and be called the interruptive command that calls supervisory routine (SVC), local management program 16-1 just handles then.
The order format of instruction GETMAIN as shown in Figure 3.Symbol 31 expression operational codes, symbol 32-35 represents operand field.Field 32 comprises the required size of using the zone, and field 33 comprises the type that this is regional, and field 34 comprises the acknowledgement field domain addresses that wherein has the region allocation result, and field 35 comprises other operand.Area type is with 0-255 totally 256 numbers numbering, and each numbering all has the implication different with other numbering.Represent the zone in the global resource 14 0-127 number, represent the zone in the local resource 19 128-255 number.Whether the type of local management program 16-1 Examination region field 33 asks to distribute regional area (101) with decision.If conclusion is a "Yes", just specify the zone (102) of local resource 19 according to request.And then control is got back on the application program 18.If the conclusion of step 101 is "No", just local management program 16-1 asks global administration's program 13 to carry out (111) so.This is called as global administration's routine call.
The handling interrupt of local cpu 3-1 this moment in regional area 12-1 carried out, and handles but carry out in other regional area 12.
Earlier lay down a definition as follows now to the control block 15 that is used for above-mentioned processing.Program block 15 is positioned at global resource 14, and its form as shown in Figure 4.It comprises the various control informations in each regional area.Row 41 explanation regional area numbers, row 42 are presented at regional area and carry out several overall CPU(ASCBGCPU that handle), row 43 show a plurality of local cpus (AsubLCPU) of just carrying out processing at regional area, row 44 show a plurality of can carrying out of tasks (ASCBTCBS) in regional area, row 45 show other control information.
Local management program 16-1 files a request to the firmware 21 of carrying out the local cpu 3-1 that handles at regional area, searches other regional area (112) that ready task is arranged and do not carry out.Firmware 21 carries out in processing shown in Figure 6 under microprogram control.At first check the row 42(ASCBGCPU in first regional area of control block 15) and row 43(ASCBLCPU) whether be " 0 " (301).If be " 0 ", just check 44(ASCBTCBS) whether greater than " 0 " (302).If greater than " 0 ", just the number of current regional area is reported to local management program (303).
If the conclusion of any one step of step 301 and 302 li is " non-", so just whether check checked last regional area (304) of control block 15, if do not check, so just check next regional area (305), and repeating step 301 and 302.If the conclusion of step 304 is "Yes", just report the zone (306) that does not have with local management program correspondence so.
Local management program 16-1 detects replying of firmware 21, and whether decision exists corresponding regional area (113).If any, just propose the number of regional area, request firmware 21 is controlled, and regional area (114) is transferred in processing.Firmware 21 responds in view of the above, handles, and its processing procedure as shown in Figure 7.It at first checks the row 42(ASCBGCPU of regional area) whether be " 0 " (401), row 43(ASCBLCPU) whether be " 0 ".If the conclusion that both have is a "No", just notify local management program CC ≠ 0(405).If the conclusion of step 401 and 402 all is " 0 ", just to the row 43(ASCBLCPU of regional area (403)) set, and to local management program (404) report CC=0.Whether local management programming check CC equals 0(115), if were 0, would ask firmware that CPU is discharged from current regional area, and conversion regional area (116).The conversion in the space of being handled by CPU is at United States Patent (USP) the 4th, 326, announces on 248.It will be replaced by new address with a content that is called the register of segment table original address, thereby can carry out the processing of new address area.
The processing of step 111 in the present key drawing 2 again.When calling global administration's program, the content of program status word (PSW) (PSW) control register (comprising the segment table original address) and general-purpose register is stored in the prefix area of the CPU of the processing that beginning is new, as common supervisor call.In the present embodiment, because many overall CPU are arranged, so the content of these registers is stored in the prefix area of overall CPU use, overall CPU is by carrying out this processing to the response of step 111.Select an execution this moment from non-busy overall CPU.Were it not for available overall CPU, so just queuing in order.
When global administration's program 13 is received global administration's routine call, just carry out processing shown in Figure 5.Global administration's program is blocked the global resource storage area, prevents that it from being used (201) by other processing.Whether in order to check the local cpu of doing this processing to discharge from regional area, whether the regional area ASCBLCPU row 43 of global administration's program 13 check control blocks 15 are " 0 " (202).If not " 0 ", repeat above-mentioned steps.If " 0 ", so just the ASCBGCPU row 42 with regional area increase 1(203), discharge then and block (204), root pushes away the information (205) that writes in the prefix area 22 then, access regional area 12-1, and read the operand (206) of the GETMAIN order of sending, according to this information designated storage area (207), set (208) in the address location of 19 li of the specified local resources of the acknowledgement field domain addresses 34 of assignment in operand.The same with step 201 then, block storage area (209), the ASCBGCPU row 42 of control block 15 regional areas are subtracted 1(210).Discharge subsequently and block (211).
In view of the above, when the application program of the regional 12-1 in part is sent instruction, the instruction of Request Processing, when it handles regional beyond the related regional area 12-1, processing be can't help local management program 16-1 and is carried out, and is carried out by global administration's program 13, and result is offered local management program 6-1.
During step 112-116 in the local cpu of not handling any regional area carries out Fig. 2, the processing of local management program 16-1 just recovers.
In another embodiment, give priority level to regional area, if a ready task is arranged in regional area, it is current also at the regional area of handling, so current handling interrupt is handled ready task that priority is higher than local cpu.
Therefore, the local management program can not be subjected to the influence of other processing and independently handle.
Fig. 1 〈 ﹠﹠〉Fig. 2 〈 ﹠﹠〉Fig. 3 〈 ﹠﹠〉Fig. 4 〈 ﹠﹠〉Fig. 5 〈 ﹠﹠〉Fig. 6 〈 ﹠﹠〉Fig. 7 〈 ﹠﹠〉Fig. 8 〈 ﹠﹠ 〉
Claims (8)
1, a multiprocessing system with a primary memory and a plurality of processors, it is characterized in that: said primary memory has a common area and a plurality of regional area, each regional area has a local management program, said common area has global administration's program, said a plurality of processor is handled by said local management program or said global administration procedure sharing main memory, said local management program is when needing the said common area of access because of the need processing, global administration's program is carried out in request, and said global administration program is carried out processing when receiving such request.
2, multicomputer system described in the claim 1, it further is characterized as has an application program in said regional area, it is carried out down by said local management programmed control, and send processing to said local management program and ask, when said local management program is received Request Processing, can in the regional area at this local management program place, carry out if handle, then carry out and handle, if can not in regional area, carry out processing, then ask said global administration program to be carried out and handle.
3, the multicomputer system described in the claim 1, it further is characterized as in one group of said processor, and each processor is carried out the processing in the regional area, and carries out the processing that need do access in said common area by the another one processor.
4, multicomputer system described in the claim 1, its further feature are in a time, have only a said processor to operate under said local management programmed control.
5, multicomputer system described in the claim 2, it further is characterized as said local manager after request is carried out, and makes by the local management program and carries out the processor of handling, and commentaries on classics is carried out by another local management program and is handled.
6, multicomputer system described in the claim 5, it further is characterized as in said common area, a control block is arranged, whether whether be used to store has processor carry out handling for regional area to have processor just to carry out the information whether processing handling and will carry out will be carried out by the local management program in regional area in regional area, said another local management program is meant that its position breath that is stored in said control block has shown that no any processor is its execution processing, do not have the processing that any processor carried out by therein local management program and do not have one of the said supervisory routine that any processor will be carried out the said regional area of processing therein.
7, has a multicomputer system by the shared primary memory of a plurality of processors, it is characterized by: said primary memory has a common area and some regional areas, some native processors, each processor is only carried out processing among in said regional area, and an overall processor is carried out the processing that needs access in said common area.
8, one is used to control the method for multicomputer system, said system comprises that one has the primary memory of a shared zone and some regional areas, each said regional area respectively has a local management program, global administration's program for said common area use, a plurality of processors of sharing said primary memory, they or carry out by said local management program or said global administration program and to handle, said regional area respectively has an application program, said method be characterized as following operation steps: carry out said application program, when the said local management program of said application requests deals with, whether decision is handled and to be carried out in one of can be in said regional area, can in regional area, carry out if handle, the local management program makes to carry out and handles, can not in regional area, carry out if handle, ask said global administration program to be carried out and handle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN85104907A CN85104907B (en) | 1985-06-25 | 1985-06-25 | Hierarchy multprocessor system and method of control |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN85104907A CN85104907B (en) | 1985-06-25 | 1985-06-25 | Hierarchy multprocessor system and method of control |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN85104907A CN85104907A (en) | 1987-02-04 |
| CN85104907B true CN85104907B (en) | 1987-10-14 |
Family
ID=4794138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN85104907A Expired CN85104907B (en) | 1985-06-25 | 1985-06-25 | Hierarchy multprocessor system and method of control |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN85104907B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100382070C (en) * | 2003-07-24 | 2008-04-16 | 国际商业机器公司 | Cluster data port services for clustered computer system |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7139881B2 (en) * | 2003-09-25 | 2006-11-21 | International Business Machines Corporation | Semiconductor device comprising a plurality of memory structures |
| US7386687B2 (en) * | 2005-01-07 | 2008-06-10 | Sony Computer Entertainment Inc. | Methods and apparatus for managing a shared memory in a multi-processor system |
-
1985
- 1985-06-25 CN CN85104907A patent/CN85104907B/en not_active Expired
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100382070C (en) * | 2003-07-24 | 2008-04-16 | 国际商业机器公司 | Cluster data port services for clustered computer system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN85104907A (en) | 1987-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US5386566A (en) | Inter-processor communication method for transmitting data and processor dependent information predetermined for a receiving process of another processor | |
| US4796178A (en) | Special purpose processor for off-loading many operating system functions in a large data processing system | |
| JP2866241B2 (en) | Computer system and scheduling method | |
| US4779194A (en) | Event allocation mechanism for a large data processing system | |
| KR930000853B1 (en) | Database processing system using multiprocessor system | |
| US5613139A (en) | Hardware implemented locking mechanism for handling both single and plural lock requests in a lock message | |
| US5884077A (en) | Information processing system and method in which computer with high load borrows processor of computer with low load to execute process | |
| US6952827B1 (en) | User program and operating system interface in a multithreaded environment | |
| US5666523A (en) | Method and system for distributing asynchronous input from a system input queue to reduce context switches | |
| US5010482A (en) | Multi-event mechanism for queuing happened events for a large data processing system | |
| US4631674A (en) | Active wait | |
| US3665487A (en) | Storage structure for management control subsystem in multiprogrammed data processing system | |
| JP2539352B2 (en) | Hierarchical multi-computer system | |
| US3639912A (en) | Management control subsystem for multiprogrammed data processing system | |
| Manner | Hardware task/processor scheduling in a polyprocessor environment | |
| EP0362903B1 (en) | A special purpose processor for off-loading many operating system functions in a large data processing system | |
| CN85104907B (en) | Hierarchy multprocessor system and method of control | |
| EP0343646B1 (en) | Task execution control method for a multiprocessor system with enhanced post/wait procedure | |
| Thompson et al. | The D825 automatic operating and scheduling program | |
| JPH02171952A (en) | Dispatching system for multiprocessor | |
| JP2787107B2 (en) | Buffer control system and device | |
| JPH03116261A (en) | Multiprocessor control system | |
| JPH011048A (en) | Process priority control system | |
| JPS58221447A (en) | Data processor | |
| JPS5839342B2 (en) | Multi-processor system NIOKEL |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| PB01 | Publication | ||
| C06 | Publication | ||
| C13 | Decision | ||
| GR02 | Examined patent application | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CX01 | Expiry of patent term |