+

WO1999063449A1 - Procede permettant d'ameliorer la concurrence d'acces dans un systeme informatique - Google Patents

Procede permettant d'ameliorer la concurrence d'acces dans un systeme informatique Download PDF

Info

Publication number
WO1999063449A1
WO1999063449A1 PCT/US1999/012433 US9912433W WO9963449A1 WO 1999063449 A1 WO1999063449 A1 WO 1999063449A1 US 9912433 W US9912433 W US 9912433W WO 9963449 A1 WO9963449 A1 WO 9963449A1
Authority
WO
WIPO (PCT)
Prior art keywords
list
data
tasks
task
locked
Prior art date
Application number
PCT/US1999/012433
Other languages
English (en)
Inventor
Clifford L. Hersh
Herbert W. Sullivan
Original Assignee
Chopp Computer Corporation
TODD, Joan
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chopp Computer Corporation, TODD, Joan filed Critical Chopp Computer Corporation
Priority to AU43316/99A priority Critical patent/AU4331699A/en
Priority to US09/701,797 priority patent/US7111293B1/en
Publication of WO1999063449A1 publication Critical patent/WO1999063449A1/fr

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
    • G06F9/54—Interprogram communication
    • G06F9/544—Buffers; Shared memory; Pipes

Definitions

  • the present invention relates generally to multitasking computer systems, and more particularly, to multiprocessor systems and a method for increased concurrency in such systems.
  • a multiprocessing computing system may be defined as a computer assembled from a plurality of independently or semi-independently operating intelligent, i.e., processor-based, stations.
  • the stations are generally interconnected for communication by a communication bus.
  • Each processor may perform a plurality of tasks with apparent concurrency.
  • Distributing the system intelligence among a plurality of stations may improve the processing throughput of the computing system, allow the system to modularly and uniformly expand to meet increasing computing requirements, and permit greater efficiency and productivity through concurrency.
  • the invention is directed to a method for operating a computer system.
  • the computer includes at least one processor.
  • the method includes establishing a plurality of memory units each having a corresponding memory location.
  • a plurality of tasks running on the processor are executed, and the plurality of tasks are operable to share data.
  • a plurality of lists for each memory- location are defined, and at least one of said lists is locked if the data is invalid.
  • An entry is inserted into said locked list corresponding to one of said tasks.
  • the locked list is unlocked, and a determination is made if data is inputted in the memory location between the determining step and the unlocking step.
  • the invention is directed to a method for operating a computer system.
  • the system includes at least one processor.
  • the method includes establishing a plurality of memory units each having a corresponding memory location.
  • a plurality of tasks are run on said processor, and the plurality of tasks are operable to share data.
  • An entry corresponding to one of the tasks is inserted into one of the lists if the list is unlocked.
  • the method also includes determining if another of said lists is unlocked if said one list is locked.
  • the invention is directed to a method for synchronizing processes in a computer system.
  • the computer system includes at least one processor.
  • the method includes establishing a plurality of memory units each having a corresponding memory location.
  • a plurality of tasks running on the processor are executed, and the plurality of tasks are operable to share data located in the memory units.
  • a plurality of lists for each memory location are defined.
  • a list is locked if the data is not valid.
  • An entry is inserted into the locked list corresponding to one of the tasks, and the list is unlocked.
  • the entered task is suspended until valid data is found in the memory unit.
  • the method also includes the steps of reading valid data and determining if other data appears in said memory location before said locking step and after said unlocking step.
  • the other data is read if it appears in the memory unit.
  • Implementations of the invention include one or more of the following.
  • the locking step further comprises activating selected other ones of said plurality of tasks that are entered on said locked lists.
  • the plurality of lists may form a linked list.
  • the plurality of lists may be between four and eight.
  • the method may also include the step of transferring the operation of said locked list when said locked list is locked by another one of said plurality of tasks.
  • the invention is directed to a computer system having enhanced concurrency that includes a plurality of processors.
  • a plurality of tasks are run on the plurality of processors.
  • the computer system may also include a plurality of memory units each having a corresponding memory location, and a plurality of lists corresponding to each of said memory locations.
  • One of said tasks may be responsible for activating selected ones of said plurality of tasks contained on the same list as said one task.
  • FIGS. 1A - ID are flowcharts illustrating a preferred method of reading data from a mailbox.
  • FIGS. 2 A - 2D are flowcharts illustrating a preferred method of inputting data in a mailbox.
  • FIG. 3 is a flowchart of a "one last look" method.
  • a multitasking computer system may be defined as a computer system that permits multiple threads of execution in a single execution space.
  • a multiprocessing computer system may be defined as a computer system that allows multiple processing units to share data.
  • various tasks in the computer system communicate using commonly accessible "mailboxes."
  • a mailbox is an area of memory that contains a number of states, and may be identified by a virtual address. When one task produces an item of valid data that may be desired by multiple other tasks, the task holding the valid data inputs it into a mailbox. Generally, other tasks read the valid data from the mailbox.
  • the task that inputs the valid data into the mailbox is caused to notify other tasks addressing the mailbox that the data is contained therein. This means that there is no need for central coordination of mailbox accesses. Further, this minimizes busy waits for valid data. This, permits tasks and processors to allocate their resources more productively.
  • the mailbox is a memory location.
  • the mailbox includes a pointer to the address corresponding to valid data.
  • a number such as a string of zeros, is employed if the data is invalid.
  • the preferred method will be described with reference to inputting or reading data to or from the mailbox.
  • the data to be accessed can be located in the same memory system as the mailbox.
  • the term "task” refers to a set of instructions running on a processor. Thus, the processor executes the task.
  • the terms "invalid” and "valid” refer to arbitrary logical status. In the preferred method, a task reads valid data and does not read invalid data. Other states may be used such as "true” and "false”.
  • a plurality of lists of tasks wait for valid data associated with the mailbox.
  • conflicts in accessing data and delays in obtaining data can be minimized. This is because only a small number of instructions, for example, under 30, are required to read or place data to or from the mailbox. These functions occupy a small percentage of a task's activity time, such as 5% or less. Accordingly, it has been found that a smaller number of lists can eliminate most conflicts. However, a list for each active task may be required to eliminate all conflicts. Any number of lists may be used. For example, the number of lists may be between four and eight.
  • the lists are a linked list.
  • a linked list can employ locations in the same memory system as the mailboxes.
  • the linked list includes locations having suitable space for at least two pointers.
  • the locations in the middle of the list may contain a forward pointer to the next location including the list, and a backward pointer to the preceding location including the list.
  • the top location on the list may contain a pointer forward to the next location and a pointer backward to an illegal address, such as zero.
  • the bottom location on the list can contain a pointer backward to the preceding location and a forward pointer to an illegal address, such as zero.
  • FIGs 1 A- ID are flow charts showing a preferred method for reading data from a mailbox.
  • data in a mailbox is read (step 101).
  • the mailbox may contain a pointer to the memory location containing the desired valid data or an illegal address corresponding to invalid data, such as zero.
  • the mailbox is then addressed and interrogated to determine if the data is valid (step 102).
  • a mailbox may contain or is about to contain data associated with a program of a task seeking such data. If valid data is in the mailbox, it is read from the memory address indicated in the mailbox (step 104).
  • the task may also read the data during step 102.
  • One aspect described herein is that the task performing steps 101 and 102 contacts other tasks waiting for valid data from that mailbox.
  • FIGS. 1 A - ID illustrate lists LO and LI that represent tasks waiting to receive valid data from a mailbox
  • FIGS. 2A-2D shows lists L2 and L3 that identify tasks waiting to input valid data in the mailbox.
  • a task list is locked when it is manipulated by a task. This prevents access by other tasks to that list.
  • list LO is locked (step 106)
  • the task determines if list LO was previously locked by another task (step 108). If list L0 was previously locked by another task prior to the current task executing steps 106 or 108, the prior task operates list L0. The current task then proceeds to check list LI (step 122), as described below.
  • the current task determines if list L0 is empty (step 1 10). If the list L0 is not empty, the current task then removes the first waiting task ("waiter") from the list (step 116).
  • a waiter may be defined as a task that waits for valid data to be in a mailbox. In this configuration, a waiter may be in a suspended state, as described below.
  • the current task then unlocks list L0 such that other tasks may access the list (step 118). Additionally, the current task causes the task associated with that waiter to execute a wake code (step 120).
  • a wake code may be defined as a set of computer instructions that cause a task to activate from a suspended mode. Then, the current task again locks list L0 (step 106).
  • step 108 If list L0 was previously locked (step 108), the current task locks list LI (step 122), and then checks if that list was previously locked (step 124). If list LI was not previously locked at step 126, the current task determines if list LI is empty (step 128). If list LI is not empty, the current task removes the waiters from list LI (step 132) and causes the waiters to execute their wake codes (step 134), as described above. This process repeats until list LI is empty or the list LI has been found locked. When list LI is empty, it is unlocked (step 130). The task then continues to execute its remaining instructions as defined by the system (step 126). If list LI was previously locked by another task prior to steps 122, 124 or 126, that task operates the locked list. The current task then executes step 126.
  • list LO is locked (step 140).
  • the current task determines if list LO was previously locked (step 142). If the list was not previously locked, the current task is entered on list LO (step 144). List LO is then unlocked (step 146), and the current task checks for valid data (step 148). The current tasks then determines the validity of the data (step 150). This above process may be referred to as "one last look. " as described below. In the preferred method, this configuration keeps two or more tasks, which are synchronized by passing data through the mailbox, from missing each other.
  • step 152 the task is suspended (step 152). This releases computer system resources for other tasks, or the task can complete other system instructions. This means that the task does not busy wait or periodically check for valid data in the memory location.
  • the current task will be activated by another task which has found valid data in the memory location, or by a task which previously inputted the valid data (step 154). When activated by another task, the current task proceeds to read the valid data from the mailbox (step 104). If the data is valid (step 150) and list L0 was previously locked (steps 156 and 158), the current task is suspended until activated (steps 152 and 154). When activated by another task, the current task proceeds to read the valid data from the mailbox (step 104).
  • the current task If the data was valid (step 150) and the current task determines that list L0 was not locked (step 156 and 158), the current task reads the data. This may occur as the current task determines the validity of the data. Then, the current task executes step 202. If the current task is on the list LO at step 202, the current task is removed from the list (step 204). The next waiting task is also removed from the list LO (step 206), and list LO is again unlocked (step 208). The task then causes the other task removed from the list to execute its wake code (step 210). Next, in accordance with the preferred method, the current task determines if the list L0 was previously locked (steps 212 and 214). If so, the current task then executes other system instructions (step 220). If the list was not previously locked, the current task then determines if the list is empty (step 216).
  • step 216 If list L0 is not empty (step 216), the current task loops back to remove and notify the next task at the top of the list (steps 206 - 216). If on any such loop, the list is found locked at steps 214 or 216, the current task continues to perform other system instructions (steps 220). When the list is empty, it is unlocked (step 218) and the current task continues to perform other system instructions (step 220).
  • the current task may not be on the list L0 (step 202). This means that another task has removed the current task from the list and is operating to activate it. However, the current task also operates list L0. Thus, the current task must activate the other tasks on list L0. The current task may then suspend itself until activated by another task.
  • Steps 228 through 238 operate similar to steps 206 through 218. However, when list L0 is found locked (step 234) or empty (step 236), the current task is suspended (step 240). When activated by another task (step 242), the current task proceeds to read the valid data from the mailbox (step 104).
  • the current task proceeds as follows. List LI is locked and then the current task determines whether that list was previously locked (steps 180 and 182). If list LI was previously locked, the current task returns to lock list L0 (step 140), and continues to search for an unlocked list.
  • This example provides two possible lists. However, more lists may be used. If the number of lists equals the number of tasks, no task will repeat this loop. If list LI was not previously locked (step 182), the current task becomes a waiter on list LI (step 184), the list LI is unlocked (step 186), and a "one last look" check is made for valid data (steps 188 and 190). If the data is not valid, the current task is suspended (step 152).
  • step 190 If the data is valid (step 190), then the current task reads the data as it confirms the validity of the data. List LI is then locked (step 192), and a check is made whether it was previously locked (step 194). If list LI was previously locked (step 194), list LI is unlocked (step 196), and the current task is suspended (step 152). If the data was valid and it is determined that list LI was not locked, at 0 steps 190, 194. and 196, respectively, then the current task proceeds (step 252).
  • FIG. ID shows that if the current task is on list LI (step 252) that task is removed from the list (step 254), and also removes the next waiting task from the list LI (step 256). The task then unlocks list LI again (step 258). The task then causes the other task removed from the list to execute its wake code (step 260). 5 Next, the current task determines whether the list LI was previously locked (steps 262 and 264). If it was not previously locked, the current task determines whether the list is empty (step 266).
  • step 266 If list LI is not empty (step 266), the current task loops back to activates task at the top of the list (steps 256 through 266). If on any such loop, the list is found 0 locked at steps 264 or 266, the current task continues step (270). When the list is empty, it is unlocked (step 261), and the current task continues to execute other system instructions (step 270).
  • step 252 If the current task is not found on the list LI (step 252), this means that some other task has taken the current task off the list and is attempting to activate it. 5 The current task also operates list LI . Thus, this task must activate the other tasks on list LI before suspending itself. The current task can then be activated by another task.
  • the method of steps 278 through 288 is similar to steps 256 through 268 described above. However, when list LI is found locked (step 284) or empty (step 286), the current task is suspended (step 290). When activated by another task (step
  • the current task proceeds to read the valid data from the mailbox at step 104.
  • This preferred method of finding an available list is simple. Further, a task should not be significantly delayed in finding an unlocked list. Thus, enhanced concurrency can be achieved over prior systems.
  • FIGS. 2A-2D are flow charts showing the input of data into a mailbox in accordance with the preferred method.
  • data is read in a mailbox (step 301).
  • the mailbox is then addressed and interrogated to determine whether the data is valid (step 302). If valid data is not already in that mailbox, the mailbox is marked as not containing valid data.
  • the valid data is placed in the mailbox (step 304).
  • lists L2 and L3 identify tasks waiting to input valid data in the mailbox. This includes a current task waiting to input data in the mailbox. The current task determines whether list L2 was previously locked (step 304). If so, that list remains locked (step 306).
  • the current task determines whether list L2 is empty (step 310). If the list L2 is not empty, then the current task removes the first waiter from that list (step 316). Also, the current task unlocks list L2 so that other tasks may access it (step 318), and causes the task associated with that waiter to execute its wake code (step 320). Then, the current task locks list L2 again (step 306). The above process continues from step 306 until all tasks on list L2 seeking the valid data from the mailbox have been activated. Alternatively, this process continues until the list L2 is empty, or the list L2 has been found locked. When list L2 is found empty, it is unlocked (step 312), and the current task proceeds to list L3 (step 322).
  • step 334 If list L2 was previously locked (step 301) or list L2 was found empty (step 310) and then unlocked (step 312), the current task then locks list L3 (step 322). Then, the current task checks if list L3 was previously locked (step 324). If list L3 was not previously locked, the current task determines whether list L3 is empty (step 328). If list L3 is not empty, the current task proceeds to remove the waiters from list L3 (steps 332 and 334) and causes them to execute their wake codes (step 306). This process repeats by looping back to step 322. The process may continue until list L3 is empty or has been found locked. When list L3 is empty, it is unlocked and the current task continues to execute other system instructions (steps 330 and 338). If list L3 is found to have been previously locked at steps 322 or 324, this indicates activity by another task. The current task then executes other system activities (step 338). The current task also defers to the other task to activate the other tasks listed on list L3.
  • the list L2 is locked (step 340). The current task also checks whether the list was previously locked (step 342). If list L2 was not previously locked, the current task is inserted onto list L2 (step 344). List L2 is then unlocked (step 346)and a "one last look" check is made for valid data (steps 348 and 350), as described below. If the data is still invalid, the current task is suspended (step 352). The current task is then activated by another task that has found or inputted valid data in the memory location (step 354). If the data in the mailbox is valid and list L2 was previously locked (steps
  • the current task is suspended until activated (steps 352 and 354).
  • the current task proceeds to input valid data into the mailbox (step 304).
  • step 400 If the data was not valid (step 350) and list L2 was not previously locked (steps 356 and 358), the current task continues to execute other system commands (step 402).
  • the current task determines whether the list L2 was previously locked (steps 412 and 414). If so, the current task proceeds to input valid data in the mailbox (step 304). Otherwise, the current task determines whether the list is empty (step 416). If list L2 is not empty (step 416), the current task loops back to remove and activate the next task at the top of the list (steps 406 - 416).
  • the current task inputs valid data in the mailbox (step 304).
  • the list is empty, it is unlocked (step 418).
  • the current task then inputs valid data in the mailbox (step 304).
  • the current task may be removed from the list and is being activated by another task.
  • the current task also operates list L2.
  • the current task must activate the other tasks which are on list L2 before suspending itself.
  • the current task may then be activated by another task.
  • the method of steps 428 through 438 is similar to steps 406-418. However, when list L2 is found previously locked (step 434) or empty (step 436), the current task is suspended (step 440). When activated by another task (step 442), the current task proceeds to input valid data in the mailbox (step 304).
  • the current task locks list L3 (step 380) and then determines whether that list was previously locked (step 382). If list L3 was previously locked, the current task loops back to attempt to access list L2 (step 340), and continues to find an unlocked list.
  • This example provides two possible lists. However, more than two lists may be employed.
  • step 382 If list L3 was not previously locked (step 382), the current task is inserted on list L3 as a waiter (step 384). The list L3 is also unlocked (step 386). and a "one last look" check is made for valid data (steps 388 and 390), as described below. If the data is valid, the current task is suspended (step 352).
  • step 392 If the data is not valid (step 390), list L3 is locked, (step 392) and a check is made whether list L3 was previously locked (step 394). If list L3 was previously locked, list L3 is unlocked (step 396), and the current task is suspended (step 352). If the data was not valid and the list L3 was not previously locked (steps 390, 394, and 396), the current task proceeds to step 452. If the current task is on list L3 (step 452), it is removed from that list (step 454). Also, the next waiting task from the list L3 is removed (step 456), and the current task unlocks list L3 again (step 458). The current task then causes the other task removed from the list to execute its wake code (step 460). The current task then determines whether the list L3 was previously locked (steps 462 and 464). If it was not previously locked, the current task determines whether the list is empty (step 466).
  • list L3 is not empty, the current task loops back to remove and activate the next task at the top of the list (steps 456-466). If on any such loop, the list is locked (steps 464 and 466), the current task inputs valid data in the mailbox (steps 304). When the list is empty, it is unlocked (step 468), and the current task determines if valid data is inputted in the mailbox (step 304).
  • step 452 If the current task does not find itself on list L3 (step 452), this may be because some other task has taken the task off the list and is trying to activate it.
  • the current task operates list L3. Thus, the current task must activate other tasks on list L3 before suspending itself.
  • the current task then may be activated by another task.
  • the method of steps 478 through 488 is similar to steps 456 through 468. However, when list L3 is found previously locked (step 484) or empty (step 486), the current task is suspended (step 490). When activated by another task (step 492), the current task proceeds to input valid data in the mailbox (step 304).
  • the term "one last look" refers to the situation where a task that has placed itself on a list for valid data from a mailbox receives that data.
  • the preferred method allows tasks addressing the same mailbox to be coordinated. This means that a current task inputting valid data in a mailbox will activate another task that was entered onto a wait list, while the data was placed in the mailbox by the current task. Accordingly, a current task may not input valid data in the mailbox, while another task is inserted on a list to receive that data.
  • the preferred method avoids the potentially infinite period of time that a task may sleep until another task activates it. Thus, delays in system processing time may be minimized.
  • FIG. 3 shows that a task first looks at the data in the selected mailbox and determines whether it is valid data (steps 548 and 550). If the data is valid, it reads the data and continues with other system activities (steps 551 and 553).
  • step 555 the task is inserted onto a wait list (step 555). Then, the "one last look” determines if the data is valid (steps 557 and 559). This occurs because, while the task that may have located an unlocked list, locked it, placed itself on the list and unlocked the list again, another task may have put valid data in the mailbox. Without this preferred "one last look” method, activating the inserted task may be delayed until another task seeks valid data from the mailbox.
  • the task will suspend until activated by another task, such as the next task inputting valid data in the mailbox (steps 552 and 554). This method corresponds generally to steps 352 and 354.
  • the task can take itself off the list (step 561). Alternatively, the task may read the data and continue its other system activities. This means that the task may assume that the data was initially found in the mailbox (steps 563 and 553).

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multi Processors (AREA)

Abstract

L'invention concerne un procédé permettant d'améliorer la concurrence d'accès dans un système informatique multiprocesseur. Différentes tâches communiquent dans un système informatique au moyen de boîtes à lettres pour accéder à des données valables issues de l'intérieur de la boîte à lettres. Une tâche contenant les données valables place ces données dans une boîte à lettres et d'autres tâches lisent les données valables issues de la boîte à lettres. La tâche qui introduit les données valables dans la boîte à lettres est amenée à informer d'autres tâches s'adressant à la boîte à lettres que les données valables y sont contenues. Par conséquent, il n'est pas nécessaire de mettre en oeuvre une coordination centrale de l'accès à la boîte à lettres. En outre, les files d'attente sont diminuées et les ressources des tâches et des processeurs sont utilisées avec plus d'efficacité. La coordination de plusieurs listes associées à chaque boîte à lettres permet de diminuer les risques de conflits liés à l'accès aux données et les risques de retard liés à l'obtention des données.
PCT/US1999/012433 1998-06-03 1999-06-03 Procede permettant d'ameliorer la concurrence d'acces dans un systeme informatique WO1999063449A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
AU43316/99A AU4331699A (en) 1998-06-03 1999-06-03 Method for increased concurrency in a computer system
US09/701,797 US7111293B1 (en) 1998-06-03 1999-06-03 Method for increased concurrency in a computer system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US8789198P 1998-06-03 1998-06-03
US60/087,891 1998-06-03

Publications (1)

Publication Number Publication Date
WO1999063449A1 true WO1999063449A1 (fr) 1999-12-09

Family

ID=22207885

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US1999/012433 WO1999063449A1 (fr) 1998-06-03 1999-06-03 Procede permettant d'ameliorer la concurrence d'acces dans un systeme informatique

Country Status (2)

Country Link
AU (1) AU4331699A (fr)
WO (1) WO1999063449A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2003038614A3 (fr) * 2001-10-30 2004-07-22 Koninkl Philips Electronics Nv Procede et systeme permettant de garantir la consistance sequentielle dans des calculs repartis
WO2004051466A3 (fr) * 2002-12-02 2005-08-11 Globespan Virata Inc Procede et systeme de messagerie entre taches entre plusieurs processeurs
KR101280805B1 (ko) 2011-06-01 2013-07-05 건국대학교 산학협력단 태스크 기반의 동시성 제어 장치 및 방법

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4707781A (en) * 1979-01-09 1987-11-17 Chopp Computer Corp. Shared memory computer method and apparatus
US5210828A (en) * 1988-12-29 1993-05-11 International Business Machines Corporation Multiprocessing system with interprocessor communications facility
US5438680A (en) * 1988-04-29 1995-08-01 Intellectual Properties And Technology, Inc. Method and apparatus for enhancing concurrency in a parallel digital computer

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4707781A (en) * 1979-01-09 1987-11-17 Chopp Computer Corp. Shared memory computer method and apparatus
US5438680A (en) * 1988-04-29 1995-08-01 Intellectual Properties And Technology, Inc. Method and apparatus for enhancing concurrency in a parallel digital computer
US5210828A (en) * 1988-12-29 1993-05-11 International Business Machines Corporation Multiprocessing system with interprocessor communications facility

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2003038614A3 (fr) * 2001-10-30 2004-07-22 Koninkl Philips Electronics Nv Procede et systeme permettant de garantir la consistance sequentielle dans des calculs repartis
WO2004051466A3 (fr) * 2002-12-02 2005-08-11 Globespan Virata Inc Procede et systeme de messagerie entre taches entre plusieurs processeurs
KR101280805B1 (ko) 2011-06-01 2013-07-05 건국대학교 산학협력단 태스크 기반의 동시성 제어 장치 및 방법

Also Published As

Publication number Publication date
AU4331699A (en) 1999-12-20

Similar Documents

Publication Publication Date Title
US7111293B1 (en) Method for increased concurrency in a computer system
US6772255B2 (en) Method and apparatus for filtering lock requests
KR100437704B1 (ko) 공간효율적인오브젝트로킹을위한시스템및방법
AU731871B2 (en) Method and apparatus for thread synchronization in object-based systems
US6247025B1 (en) Locking and unlocking mechanism for controlling concurrent access to objects
JP2866241B2 (ja) コンピュータシステムおよびスケジューリング方法
EP0747815B1 (fr) Procédé et dispositif pour éviter des boucles sans fin pendant la sérialisation de l'accès à des ressources non-sécurisés
EP0428006B1 (fr) Système et procédé de verrouillage à plusieurs niveaux
US6269391B1 (en) Multi-processor scheduling kernel
EP0533805B1 (fr) Procede de gestion efficace de la memoire centrale non virtuelle
US6910212B2 (en) System and method for improved complex storage locks
US6173442B1 (en) Busy-wait-free synchronization
CN100407136C (zh) 使用睡眠-唤醒机制来执行指令的方法和装置
KR100843548B1 (ko) 공유형 리소스의 액세스 방법과 그 장치 및 컴퓨터 판독가능한 저장 매체
US20020029239A1 (en) Method and system for enhanced concurrency in a computing environment
US5742830A (en) Method and apparatus for performing conditional operations on externally shared data
US20090235034A1 (en) Computer architecture and method of operation for multi-computer distributed processing with synchronization
EP0682312A2 (fr) Mécanisme de verrouillage implémenté en hardware pour un système d'ordinateur
EP1031925B1 (fr) Traitement coopératif de tâches dans un système d'ordinateur multithread
EP1163581B1 (fr) Conversion de processus de controle dans un systeme informatique a multiples fils d'execution
McKenney Selecting locking primitives for parallel programming
US6976260B1 (en) Method and apparatus for serializing a message queue in a multiprocessing environment
Stone Database applications of the fetch-and-add instruction
Gidron et al. Salsa: scalable and low synchronization numa-aware algorithm for producer-consumer pools
WO1999063449A1 (fr) Procede permettant d'ameliorer la concurrence d'acces dans un systeme informatique

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG US UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW SD SL SZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
WWE Wipo information: entry into national phase

Ref document number: 09701797

Country of ref document: US

点击 这是indexloc提供的php浏览器服务，不要输入任何密码和下载