CN1181434C - Adaptive Dynamic Memory Management Method - Google Patents

Abstract

The present invention discloses a self-adaptive dynamic memory management method which takes page clusters as units to allocate memory from an idle memory region for application. Pages are contained in the page clusters; the length of each page can accommodate a segment with maximal byte number. In the page clusters of the same type, the byte numbers of segments in all pages are same. When all of the segments in the page clusters are idle, the page clusters can be converted into page clusters of required type to be applied. The present invention realizes the dynamic management of memory spaces and fully utilizes memory resources.

Description

Adaptive Dynamic Memory Management Method

technical field

The invention relates to a memory management method, in particular to a memory management method in a data communication device.

Background technique

In various data communication devices, the memory space is divided into multiple fragments according to the number of bytes. The length of various fragments, that is, the number of bytes, includes 32 bytes, 64 bytes, 128 bytes, 256 bytes, and 512 words. section, 1024 bytes, 2048 bytes, 4096 bytes, etc. The requirements for various types of memory are constantly changing in the operation of equipment. In order to perform memory application and release operations at high speed, the application for the same type of memory is usually applied in batches, that is, the number of various fragments is determined at the time of system compilation. After entering the running process, this number will not be able to Dynamically adjust according to actual needs. For example, we estimate based on experience that a certain device usually needs about 50,000 slices of 32-byte memory slices, so when the system is initialized, a one-time application can store 50,000 slices of 32-byte memory area for access. During system operation, when a 32-byte memory needs to be requested, a free 32-byte memory slice is taken from this memory area for use. If there is no free memory slice in the memory area, the system returns The application fails to prompt, even if there is still a lot of free memory space in other types of memory areas at this time. Although this fixed-allocation memory management method is easy to implement, idle memory resources cannot be fully utilized during use, resulting in waste.

Contents of the invention

The object of the present invention is to provide a self-adaptive dynamic memory management method to address the deficiencies in the prior art, to dynamically allocate memory of various slice types according to actual needs during operation, so as to make full use of memory resources.

In order to achieve the above object, the technical solution adopted by the present invention is: a kind of self-adaptive dynamic memory management method, divides the memory space into different types of page clusters according to the number of bytes of various fragments, and takes the page cluster as a unit from the entire A certain proportion of the memory space is allocated in the free memory area for application. Each page cluster contains at least one page, and each page can only accommodate one fragment. In the same type of page cluster, the fragments in all pages The number of bytes of the slice is the same. When applying for a certain number of bytes of memory, the following steps are included:

A. Check whether there are free fragments in the page cluster of this byte fragmentation type, if yes, go to step E; if not, continue to step B;

B. Determine whether the ratio of the free memory area to the entire memory space is less than or equal to the set value, if yes, go to step D; if not, go to step C;

C. Divide a page cluster of a corresponding type from the free memory area, and turn to step E;

D. Find a page cluster in which all fragments are free from other types of page clusters, and change it to a page cluster of the required type; continue with step E;

E. Allocate fragments from page clusters.

Since the present invention adopts the above technical solution, the memory space is allocated in units of page clusters for applications. Page clusters contain pages, and each page can only accommodate one fragment. In actual operation, when a certain type of All the shards in the page cluster are occupied, and when the user applies for this type of shard, there are two ways to solve the problem: (1) Divide the required type of page cluster from the free memory space, and then Allocate the fragments therein; (2) Find out the page clusters in which each fragment is free from other types of page clusters, change its type, that is, change the length of the fragments to the required type, and then allocate. Since the length of the larger fragment is an integer multiple of the length of various fragments smaller than it, and each page can only accommodate one fragment, and the page cluster is composed of pages, so various types of page clusters can be convert each other. Through the division and mutual conversion of page clusters, the dynamic management of memory space is realized, and applications will not be rejected as long as there is free time, so that memory resources can be fully utilized.

Detailed ways

The technical solutions of the present invention will be further described below in conjunction with the embodiments.

The core of the technical solution of the present invention is to make a new division of the memory space, so that the application and release of the memory by the user are carried out in the divided intervals, and this new division enables the mutual conversion of different types of intervals. Various types of memory do not need to reserve a fixed amount, and are dynamically allocated according to the application during operation.

When applying for a certain number of bytes of memory, the following steps are involved:

A. Check whether there are free fragments in the page cluster of this byte fragmentation type, if yes, go to step E; if not, continue to step B;

B. Determine whether the ratio of the free memory area to the entire memory space is less than or equal to the set value, if yes, go to step D; if not, go to step C;

C. Divide a page cluster of a corresponding type from the free memory area, and turn to step E;

D. Find a page cluster in which all fragments are free from other types of page clusters, and change it to a page cluster of the required type; continue with step E;

E. Allocate fragments from page clusters.

The unit of memory to be used from the free memory area is the page cluster, while the unit of applying for memory is the unit of fragment. The function of step A is to check whether there is a fragment of the same type that has not been used in the corresponding page cluster each time you apply. Apply. If there is, allocate the fragment directly; if not, you need to take another page cluster of this type.

There are two ways to get a new page cluster, the first is to divide a corresponding type of page cluster from the free memory area, the second is to find a free page cluster from other types of page clusters, and then change it For the required type of page cluster, the so-called changing the type of page cluster is to change the type of fragments in the page cluster. For example, a page cluster containing 24 512-byte fragments can be converted into a page cluster containing 6 2048-byte fragments. It can also be converted into a page cluster containing 96 128-byte fragments. In order to reduce the shock that may be caused by page cluster type conversion, a condition is stipulated in step B as the basis for choosing the above two methods, that is, to judge whether the proportion of free memory area to the entire memory space is less than or equal to the set minimum value, If yes, choose the second option; if no, choose the first option. If the ratio is too high, the system will perform more page cluster type conversion operations, and if the ratio is too low, it will not be conducive to handling a large number of certain types of shards. According to the comparison, the minimum value of the ratio of the free memory area to the entire memory space can be set to any value greater than 10% and less than 15%.

The present invention introduces page clusters as the unit for dividing memory space. Each page cluster contains at least one page, and each page can and can only accommodate one fragment. In the same type of page cluster, the fragments in all pages The number of bytes is the same. The number of pages in each page cluster can be set arbitrarily, and in practical applications, 2 to 4 pages are suitable. Assuming that the maximum fragment length in practical applications is 4096 bytes, and the number of pages in each page cluster is 3, then the page cluster can be of the type containing 6 fragments of 2048 bytes or 12 fragments of 1024 bytes. The type of slice, etc., and so on.

In use, page clusters can be managed in the form of a linked list, and a page cluster header is set for each page cluster to record the following information:

(1) Page cluster type - the fragment type contained in this page cluster, such as 32 bytes, 64 bytes, 128 bytes, etc.;

(2) Number of pages - the number of pages contained in this page cluster;

(3) Free pointer—when the current page cluster is free, it points to the page cluster head of the next free page cluster of the same type.

The page cluster head can also carry other information used for algorithm control.

In the implementation of the present invention, the idle fragments of the same type can also be managed in the form of a linked list, and the idle fragment header pointer is set to point to the head of the idle fragment chain. The idle fragments in the chain belong to multiple different page clusters. In order to facilitate the recycling and conversion of the idle page clusters, we need to restrict the connection sequence of the idle fragments, that is, to make the pages in the same page cluster The free fragments are connected to the connected segments in the above linked list, so the fragments belonging to the same page cluster can be identified, so that when the fragments are released, they can be classified into the page clusters to which they belong. Then set the head pointer for pointing to the first fragment in this page cluster and the tail pointer for pointing to the last fragment in this page cluster for each page cluster, so that when the fragment is released, if the fragment If there is already a free fragment in the page cluster where the fragment is located, add the fragment to the tail pointer, and then adjust the tail pointer to point to the newly released fragment; Both the head pointer and the tail pointer point to this fragment, and then add this fragment to the head of the free fragment chain of the length pointed to by the head pointer. If after a fragment is released, all the fragments in the page cluster where this fragment is located become free fragments, then this page cluster becomes a free page cluster.

When a fragment application is made from a certain type of page cluster, if the free fragment header pointer is not empty, the fragment it points to is allocated to the application, and the free fragment header pointer points to the next free fragment; If the free fragment header pointer is empty, it means that this type of page cluster has been fully used, then allocate a page cluster of the same type first, and then allocate fragments from this new page cluster.

The system can also reserve a part of the memory space that does not use the method of the present invention for allocating memory applications larger than the largest fragment.

Claims (6)

1. An adaptive dynamic memory management method, characterized in that: the memory space is divided into different types of page clusters according to the number of bytes of various fragments, and the page cluster is taken as a unit from the free memory that accounts for a certain proportion of the entire memory space Area allocation for application, each page cluster contains at least one page, each page can only accommodate one fragment, in the same type of page cluster, the number of bytes of fragments in all pages is the same, When applying for a certain number of bytes of memory, the following steps are included: A. Check whether there are free fragments in the page cluster of this byte fragmentation type, if yes, go to step E; if not, continue to step B; B. Determine whether the ratio of the free memory area to the entire memory space is less than or equal to the set value, if yes, go to step D; if not, go to step C; C. Divide a page cluster of a corresponding type from the free memory area, and turn to step E; D. Find a page cluster in which all fragments are free from other types of page clusters, and change it to a page cluster of the required type; continue with step E; E. Allocate fragments from page clusters. 2. The adaptive dynamic memory management method according to claim 1, wherein the number of pages included in the page cluster is 2, 3 or 4. 3. The adaptive dynamic memory management method according to claim 1, characterized in that: setting the minimum ratio of the free memory area to the entire memory space as any value greater than 10% and less than 15%. 4. The self-adaptive dynamic memory management method according to claim 1, 2 or 3, characterized in that: the fragments belonging to the same page cluster are identified, so that when the fragments are released, they are classified into the corresponding page clusters. page clusters. 5. The adaptive dynamic memory management method as claimed in claim 4, characterized in that: for each page cluster, a head pointer for pointing to the first fragment in the page cluster and a header pointer for pointing to the last slice in the page cluster are set. A shard tail pointer. 6. The adaptive dynamic memory management method according to claim 5, characterized in that: when a fragment is released, if the page cluster where the fragment is located already has a free fragment, the fragment is added to the tail pointer After that, adjust the tail pointer to point to the newly released fragment; if there is no free fragment in the page cluster where the fragment is located, point both the head pointer and the tail pointer of the page cluster to this fragment, and then add this fragment to the head The head of the free shard chain of the length pointed to by the pointer.