RU2480819C2 - Method of optimising work with linked lists - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: method of providing read processes continuous access to a linked list which contains one write process and at least one read process and in which upon deleting the last list element: all read processes obtain access to the list without waiting and are divided into read processes which started work with the list before the write process switched to the last list element, and read processes which started work with the list after the write process switched to the last list element, wherein: a) read processes which started work with the list before the write process switched to the last list element are executed for the number of list elements containing the last element; b) read processes which started work with the list after the write process switched to the last list element are executed for the number of list elements which do not contain the last element; wherein the write process deletes the last list element only after all read processes which started work with the list before the write process switched to the last element complete work with the list.

EFFECT: optimising work of write and read processes with linked list by providing read processes continuous access to the list while maintaining integrity of the data obtained by read processes.

6 cl, 7 dwg

Description

Technical field

The invention relates to methods for fetching, addressing or distributing data in memory systems or architectures and, in particular, to methods for optimizing work with linked lists.

State of the art

All computer systems have limited resources and, in particular, limited memory. For this reason, if several processes are carried out in the system, these processes have to compete for memory. For the same reason, processes have to compete for software resources. Examples of software resources include shared libraries and various data structures.

One example of data structures is a linked list. A linked list refers to a data structure consisting of elements, each of which contains both its own data and one or two links to the next and / or previous list item. All processes working with linked lists are conditionally divided into write processes and reading processes. Recording processes include those processes that make changes to the linked list, in particular adding new items, deleting or changing existing items in the linked list. Reading processes include processes designed to retrieve data from linked list items.

Figure 1 shows a structural diagram of a variant of implementation of a linked list. The linked list 100 consists of list items 110, 120, 130, 140 and 150. Each linked list item contains its own data and a link to the next list item. So, element 110 contains a block of own data 111 and a link 112 to the next element of the list 120. Element 120 contains a block of own data 121 and a link 122 to the next element of the list 130. Element 130 contains a block of own data 131 and a link 132 to the next element of the list 140. Element 140 contains an own data block 141 and a link 142 to the next element of the list 150. The last element of the list 150 contains only the own data block 151.

An example of using a linked list is the list of trusted processes and threads used by antivirus software to identify processes and threads for which malware testing is not required.

With the simultaneous operation of the writing and reading processes with the linked list, a situation may arise when the reading process receives data from the linked list item at the time it is changed by the write process, which ultimately leads to a violation of the integrity of the received data. In order to avoid this situation, it is necessary to synchronize the reading and writing processes.

In classical schemes, the synchronization problem is solved by using mutual exclusion. The essence of using mutual exclusion is that at a certain point in time, only one type of process can work with each element of the linked list — either writing or reading. Depending on the conditions of the task, the priority in using mutual exclusion may belong to one of two types of processes.

Figure 2 shows the structural diagram of the processes of reading and writing with a linked list. The reading processes 210 and 220 read the elements 110 and 120 of the linked list 100, respectively. The write process 230 changes the element 130 of the linked list 100. In order to ensure the integrity of the data received by the reading processes 210 and 220, these reading processes must wait for the completion of the write process 230 with the linked list element 130 before starting to read this element. Thus, the priority in using mutual exclusion in the case of FIG. 2 belongs to the recording process.

However, the method of using mutual exclusion has one significant drawback, namely, the likelihood of a situation when a process of one of the types expects the completion of the process of another type a significant amount of time. Moreover, in some implementation schemes, the speed of a computer system depends on the speed of the processes with linked lists. So, for example, it happens when a linked list is used by antivirus software and contains information about trusted processes and threads, for which malware testing is not required. In this case, when each new process or thread occurs in the system of an existing process, the anti-virus software checks for the content of information about the specified processes or threads in a linked list before allowing them to be executed. If the priority in using mutual exclusion belongs to the recording processes, the execution of all processes and threads in the system may be delayed for the duration of the recording process with the linked list. If the priority in using mutual exclusion belongs to the reading processes, the possibility of making operational changes to the linked list will be lost.

Therefore, to work effectively with linked lists, it is necessary to use an approach that will allow us to avoid these restrictions and which preserve the integrity of the data obtained by reading processes.

US Pat. No. 7,536,428 describes the idea of competitive access to a linked list of read and write processes based on the use of three sublists: a read list, a write list, and a change list. All changes are made by the writing process to the change sublist, while reading processes have unlimited access to the reading sublist. Migration of changes from the change sublist to the read sublist occurs only if the reading processes do not work with the read sublist. This idea does not ensure the speed of the entry into force of changes to the linked list, since these changes take effect only at the moment when no reading process works with the sublist.

US Pat. No. 6,304,924 describes the idea of organizing access to a linked list of read and write processes in a cyclic queue. When working with a linked list item, each of the two types of processes blocks the next part of the list following the item for processes of another type. Thus, in this idea, reading and writing processes have limited access to the list, since they are forced to wait for the completion of the work of other types of processes with a linked list element.

In patents US 7117502 and US 6760726 describes the idea of organizing competitive access of read and write processes to a linked list, which combines one common drawback, namely the use of synchronized mutual exclusion operations.

In the patent US 6360220 and the application US 20020138706 describes the idea of sharing a data structure using a counter of reading processes to determine the point in time when the structure does not work with any reading process. At the point in time when reading processes do not work with the data structure, it is possible to make changes without risking loss of data integrity. Thus, in this approach, the priority of using mutual exclusion is given to reading processes, which leads to the impossibility of prompt changes.

Thus, although the listed approaches are aimed at solving certain problems in the field of optimizing work with linked lists, they have a number of drawbacks that were mentioned above. The present invention allows to more efficiently and effectively solve the problem of providing the reading processes with continuous access to the list, which preserves the integrity of the data obtained by the reading processes.

Disclosure of invention

The present invention is intended to optimize work with linked lists.

The technical result of the present invention is to optimize the operation of the writing and reading processes with linked lists by providing the reading processes continuous access to the list while maintaining the integrity of the data obtained by the reading processes for the case when one writing process and at least one reading process work with the list.

In a way to provide reading processes with continuous access to a linked list that contains one writing process and at least one reading process, when a list item is deleted, all reading processes without waiting get access to the list and are divided into reading processes that started working with the list before the writing process went to the last element of the list, and to the reading processes that started working with the list after the writing process went to the last element of the list. In this case, the reading processes that started working with the list before the writing process moved to the last element of the list are performed for the number of list elements containing the last element. Reading processes that started working with the list after the writing process has moved to the last element of the list are performed for the number of list elements that do not contain the last element. In this case, the writing process deletes the last element of the list only after all the reading processes that started working with the list before the writing process moved to the last element have finished working with the list.

In the above method, the recording process moves the item to be deleted to the end of the list if the item to be deleted is not the last in the list.

In addition, in the indicated method, the recording process also logically deletes the list item by marking the item deleted, while the data of the logically deleted list item is ignored by the reading process, and the item remains in the list.

In a method of providing reading processes continuous access to a linked list that contains one writing process and at least one reading process, when a new list item is added, all reading processes without waiting gain access to the list and are divided into reading processes that started with list before adding a new element, and on reading processes that started working with the list after adding a new element. At the same time, reading processes that started working with the list before adding a new element are performed for the number of list elements that do not contain the element to be added. Reading processes that started working with the list after adding a new element are performed for the number of list elements containing the added element.

In the indicated method, the recording process adds a new element to the end of the list if there is no logically deleted element in the list. If a logically deleted element is present in the list, the recording process adds a new element by copying the data of the new element to the logically remote element and removing the mark from the logically deleted element as deleted.

Brief Description of the Drawings

The accompanying drawings are included to provide a further understanding of the invention and form part of this description, show embodiments of the invention, and together with the description serve to explain the principles of the invention.

The claimed invention is illustrated by the following drawings, in which:

Figure 1 shows a block diagram of an embodiment of a linked list.

Figure 2 shows the structural diagram of the processes of reading and writing with a linked list.

Figure 3 shows a block diagram of an embodiment of a method for optimizing work with linked lists.

Figure 4 shows a structural diagram of a method for operating a reading process in a method for optimizing work with linked lists.

5 shows a structural diagram of a method of operating a recording process in a method for optimizing work with lists when performing a logical deletion operation of a list item.

6 shows a structural diagram of a method of operating a recording process in a method for optimizing work with lists in an operation of adding a list item.

7 shows a structural diagram of a method of operating a recording process in a method for optimizing work with lists in a delete operation of a list item.

Description of Embodiments

The objects and features of the present invention, methods for achieving these objects and features will become apparent by reference to exemplary embodiments. However, the present invention is not limited to the exemplary embodiments disclosed below, it can be embodied in various forms. The above description is intended to help a person skilled in the art for a comprehensive understanding of the invention, which is defined only in the scope of the attached claims.

Figure 3 shows one embodiment of a method for optimizing work with linked lists. This embodiment includes a list management system 300 and a linked list 100 itself. The list management system 300, in turn, consists of a switching device for an active counter 310, two reading process counters 320 and 330, and a storage device for the number of list items 340.

Each of the two reading process counters 320 and 330 is designed to count the number of reading processes that are currently working with the list. Counting is done by increasing the counter value by one if the reading process has started working with the list, and decreasing by one the counter value if the reading process has finished working with the list, with the initial counter value equal to zero. At any given time, only one of two counters can be active, that is, counting the number of readers who started working with a list of reading processes by increasing their value by one. At the same time, when a certain reading process completes work with the list, decreasing its value by one produces the counter that was active when this reading process started with the list. Thus, the value of a certain counter equal to zero will indicate that all the reading processes that started working with the list in the period of time when this counter was active have finished working with the list.

In one embodiment of the invention, instead of each counter, it is possible to use a separate pair of counters. In this case, one counter from a pair increases its value by one if the reading process has started working with the list, and the second counter from a pair increases its value by one if the reading process has finished working with the list. Thus, when all the reading processes that started working with the list in the period of time when a certain separate pair of counters were active, finish working with the list, the values of this particular pair of counters become equal to each other.

As indicated above, at any time, only one of the two counters 320 and 330 can be active. Therefore, the switching device of the active counter 310 is designed to simultaneously deactivate the active counter and activate the inactive counter.

The device for storing the number of list items 340 is designed to store and provide reading processes with a readable number of list items. The indicated readable number of list items is read in the course of working with the linked list 100. In order to ensure the integrity of the data received by the reading processes, the list of items available for reading stored in the device 340 may not include the last list item, the removal or addition of which produced at a specific point in time. In addition, as will be shown below, the reading process when starting work with the linked list 100 after increasing the value of the active counter by one receives from the device 340 the number of list items that it needs to go through. Thus, each counter also counts the number of reading processes for which the number of linked list items that were stored in device 340 during the period of time when this counter was active is available.

In the described method for optimizing work with linked lists, all the operations that the recording processes perform with the elements of the linked list 100 are divided into three types: the operation of adding a list item, the operation of logically deleting a list item and the operation of deleting a list item.

The operation of adding a list item is used by the writing process to add a new item to the linked list.

The operation to delete a list item is used by the write process to physically remove the item from the linked list. The length of the list is reduced.

When performing the operation of logical deletion of a list item, the length of the linked list does not change, since the item in this case is not physically deleted, but only marked as deleted. In addition, when performing a logical deletion operation on a list item, the data contained in the specified item remains unchanged. Thus, in the event of a situation when the data is received by the reading process from the linked list item at the time the logical deletion operation is performed with the specified item, the integrity of the received data will not be violated. If the reading process accesses the list item after the specified item has been marked as deleted, the data contained in the list item that is marked as deleted is ignored by the reading processes.

Figure 4 shows a structural diagram of the method of operation of the reading process in a method of optimizing the work with linked lists. At step 410, the reading process starts working with the linked list 100. At the same time, the value of the active counter at that moment increases by one at step 420. At step 430, the reading process receives from the storage device the number of list items 340 the number of list items that it needs to go through, after which refers to the first element of the list at step 440. At step 450, the reading process checks to see if the element to which it refers is marked as deleted. If the element is not marked as deleted, at step 460 its data is read, otherwise at 465, the data contained in the element is ignored by the reading process. After that, the reading process at step 470 checks to see if the number of elements passed by it in the linked list 100 is equal to the number of elements obtained at step 430. If this equality holds, the reading process terminates with the list at step 485, while the value of the reading process with the list of the counter active at the time of the beginning of the work decreases by one at step 490. Otherwise, the reading process goes to the next connected element by reference of the first element list 100 at step 480. Steps 450, 460, 465, 470 and 480 are repeated for subsequent elements of linked list 100 until the equality is reached.

Figure 5 shows a structural diagram of a method of operation of the recording process in a method of optimizing work with lists when performing the operation of logical deletion of a list item. At step 510, the recording process starts working with the linked list 100 and searches for the list item that must be deleted logically at step 520. If the item is not found, the recording process ends with the linked list 100 at step 540. If the item is found before completing work with linked list 100 at step 540, the recording process marks the found item deleted at step 530.

As mentioned above, the operation of logical deletion of a linked list item does not affect the integrity of the data obtained by the reading process, since when performing this operation, the recording process does not modify the data contained in the item.

Figure 6 shows a structural diagram of a method of operation of the recording process in a method of optimizing work with lists when performing the operation of adding a list item. When performing this operation, a new list item can be added to the end of the list, while the length of the list will increase, or instead of a logically deleted item, if such an item is present in the list. Adding a new element instead of a logically deleted one provides more efficient use of list space, which in turn increases the speed of search operations for list items.

At step 610, the recording process starts working with the linked list 100, after which, at step 620, it determines whether a logically deleted item is present in the list. If a logically deleted element is present in the list, the recording process copies the data of the new element to the logically deleted element at step 630, after which it removes the mark deleted from the element at step 640 and completes work with the list at step 670. Thus, all reading processes, which will then begin working with the list, will see a new element instead of a logically deleted one. If at step 620 no logically deleted item was found in the list 100, the recording process adds a new item to the end of the list at step 650, after which at step 660 it increases by one the value stored in the storage device for the number of items in the list 340, and at step 670 finishes work with the list. Thus, all the reading processes, which will then begin working with the list, will see a new element, since the equality between the number of elements passed through the reading process of the list 100 will become equal to the value stored in the storage device for the number of elements in the list 340, only after reading the new element.

Fig. 7 shows a structural diagram of a method of operating a recording process in a method for optimizing work with lists when performing a delete operation of a list item. Since in the indicated operation the element is physically deleted, the probability of obtaining non-integral data by the reading process is excluded in the considered algorithm. At step 710, the recording process starts working with the linked list 100 and, at step 720, searches for the list item to be deleted. The specified algorithm is considered for the case when the element to be deleted is found in the list. If the item cannot be found, the recording process ends with the linked list 100. After the item to be deleted is found, the recording process determines at step 730 whether this item is the last in the list. If the item to be deleted is the last in the list, the recording process in step 735 decreases by one the value stored in the storage device of the number of items in the list 340, and in step 745 switches the active counter using the switching device of the active counter 310. All processes readings, which after that start working with the list, will not see the element that needs to be deleted, since the equality between the number of elements passed by the reading process will become equal to the value stored in the device neniya number list 340 elements during the passage of the penultimate read list element process. The element that needs to be deleted can be visible only to those reading processes that started working with the list until the value of the device 340 decreased by one at step 735 and, therefore, before switching the active counter at step 745. Since these reading processes started working with the list before switching the active counter at step 745, upon completion of their work with the list, the counter will decrease by one, which after switching the active counter at step 745 will become inactive. The value “zero” of the specified counter will indicate that all reading processes that could be visible to the deleted item have finished working with the list. Therefore, in order to avoid the case when the reading process receives non-integral data, before deleting the last element of the list, the recording process needs to make sure that the value of the inactive counter is zero. To this end, at step 755, the recording process determines whether the value of the inactive counter is zero. If the value of the inactive counter is not equal to zero, the recording process at step 785 waits until this value becomes equal to zero. If the value of the inactive counter is equal to zero, the recording process deletes the last element of the list at step 765 and completes the work with the list at step 775.

If the item to be deleted is not the last in the list, the recording process marks this item as deleted at step 740 according to the algorithm for the logical delete operation described above. After that, at step 750, the recording process copies data from the subsequent element to the element, which at step 740 was marked as deleted. Next, the recording process at step 760 removes the mark deleted from the element into which data from the subsequent element was copied at step 750, and the subsequent element itself is marked as deleted at step 770. Thus, the recording process swaps the element that needs to be deleted, followed by an element. That is, the item to be deleted moves to the end of the list. Steps 740, 750, 760 and 770 are repeated until the element to be deleted becomes the last in the list, after which this element is deleted according to steps 735, 745, 755, 765, 775 and 785 of the described algorithm.

It is also worth noting that steps 745, 755 and 785, in which the active counter is switched, checking the value of the inactive counter and waiting for the time when this value becomes zero, can be performed for other operations of changing the linked list.

In conclusion, it should be noted that the information provided in the description are examples that do not limit the scope of the present invention defined by the claims. One skilled in the art will recognize that there may be other embodiments of the present invention consistent with the spirit and scope of the present invention.

Claims (6)

1. A method for providing reading processes with continuous access to a linked list that contains one write process and at least one read process, and in which, when the last element of the list is deleted:
all read processes without waiting gain access to the list and are divided into reading processes that started working with the list before the writing process went to the last element of the list, and reading processes that started working with the list after the writing process went to the last element of the list, with:
a) the reading processes that started working with the list before the writing process moved to the last element of the list are performed for the number of list elements containing the last element;
b) the reading processes that started working with the list after the writing process has moved to the last element of the list, are performed for the number of elements in the list that do not contain the last element;
at the same time, the writing process deletes the last element of the list only after all the reading processes that started working with the list before the writing process moved to the last element finished working with the list. 2. The method of claim 1, wherein the recording process moves the item to be deleted to the end of the list if the item to be deleted is not the last in the list. 3. The method according to claim 1, in which the recording process also logically deletes the list item by marking the item deleted, while the data of the logically deleted list item is ignored by the reading process, and the item remains in the list. 4. A method for providing reading processes with continuous access to a linked list that contains one writing process and at least one reading process, and in which when adding a new list item:
all read processes without waiting gain access to the list and are divided into reading processes that started working with the list before adding a new element, and reading processes that started working with the list after adding a new element, while:
a) the reading processes that started working with the list before adding a new item are performed for the number of list items that do not contain the item to be added;
b) the reading processes that started working with the list after adding a new item are performed for the number of list items containing the added item. 5. The method according to claim 4, in which the recording process adds a new element to the end of the list, if the list does not have a logically deleted element. 6. The method according to claim 4, in which the recording process adds a new element by copying the data of the new element to the logically remote element and removing the tag from the logically deleted element deleted if the logically deleted element is present in the list.

Images