CN107665146B - Memory management device and method - Google Patents

Abstract

The invention discloses a memory management device and a memory management method, and belongs to the field of computers. The memory management device manages the memory blocks and the memory fragments in a hierarchical mode, flexibly applies or releases the memory blocks according to the actual occupation condition of the memory fragments, achieves dynamic management of the memory blocks, can meet the burst of services, enhances the sharing performance of memory resources, avoids resource waste and improves the use efficiency of the memory.

Description

Memory management device and method

technical field

The present invention relates to the field of computers, and in particular, to a memory management device and method.

Background technique

With the development of computer technology and communication technology, data communication services have become more and more common in people's lives. Data communication services usually include data reception, data storage and data forwarding services. Frequent operations, specifically, when receiving data, it is necessary to apply for memory, and then store the data in the applied memory, and then forward the data after specific processing, and the memory needs to be released. It meets higher requirements, that is, less fragmentation, high efficiency, and strong anti-burst ability.

In related technologies, memory management is usually performed in a static manner, that is, the memory is configured into multiple memory blocks of the same size, and then each memory block is configured into multiple memory slices of the same length, and the memory of the same length is divided into multiple memory blocks with the same length. The memory addresses of the slices are stored in the same memory queue to maintain the allocation of memory slices in the form of memory queues. For example, when there is a business requesting memory, a free memory address is taken from the head of the above-mentioned memory queue for allocation. When the memory is released, the address of the memory to be released is stored at the tail of the above-mentioned memory queue.

In the process of realizing the present invention, the inventor found that the prior art has at least the following problems:

In the above-mentioned static memory management, a fixed number of memory blocks are set for each service. However, services are often bursty, and a service may not be able to occupy all the memory blocks occupied by a memory queue, resulting in memory loss. Waste of resources and low memory usage efficiency.

SUMMARY OF THE INVENTION

In order to solve the problems in the prior art, embodiments of the present invention provide a memory management apparatus and method.

First, the present invention provides a memory management device, the memory management device includes a plurality of functional modules, and the plurality of functional modules are used to monitor the status of at least two memory blocks; when a memory block has been fragmented, Then the state of the memory block is the occupied state, and when a memory block is not fragmented, the state of the memory block is the idle state; monitor the states of at least two memory fragments, the at least two memory fragments are in the idle state. The slice includes at least two slice lengths, and each memory slice is divided into memory blocks; if the number of free memory slices of the same length is less than the memory block application threshold, the application will be applied according to the status of the at least two memory blocks. The memory blocks in the idle state are divided; if the number of free memory fragments of the same length is greater than the memory block release threshold, the free memory fragments of the same length are released; when a service request is received, a corresponding length is allocated for the service request The free memory fragmentation; according to the memory address of the free memory fragmentation allocated by the memory fragmentation management module, the data corresponding to the service request is stored.

The memory management device provided by the present invention realizes the dynamic management of memory blocks by performing hierarchical management on memory blocks and memory slices, and flexibly applies for or releases memory blocks according to the actual occupancy of memory slices, thereby realizing the dynamic management of memory blocks and meeting business needs. The burstiness of memory resources enhances the sharing of memory resources, avoids resource waste, and improves memory usage efficiency.

Secondly, the present invention also provides a memory management method, the method specifically includes: monitoring the states of at least two memory blocks; when a memory block has been fragmented, the state of the memory block is an occupied state, when When a memory block is not fragmented, the state of the memory block is an idle state; the state of at least two memory fragments is monitored, and the at least two memory fragments include at least two fragment lengths. The shards are divided into memory blocks; if the number of free memory shards of the same length is less than the memory block application threshold, then according to the state of the at least two memory blocks, apply for the memory blocks in the free state for division; if the same length If the number of free memory fragments is greater than the memory block release threshold, the free memory fragments of the same length are released; when a service request is received, a corresponding length of free memory fragments is allocated for the service request; according to the allocated free memory fragments The memory address of the memory slice, which stores the data corresponding to the service request.

The memory management method provided by the present invention realizes the dynamic management of memory blocks by performing hierarchical management on memory blocks and memory slices, and flexibly applies for or releases memory blocks according to the actual occupancy of memory slices, and can satisfy business needs. The burstiness of memory resources enhances the sharing of memory resources, avoids resource waste, and improves memory usage efficiency.

In a possible design, the method further includes:

For a memory fragment of any length, if no application for the memory fragment of the length is received within a preset time, the memory block occupied by the memory fragment of the length is released.

Through the above memory management method, it is possible to judge whether the memory block needs to be reclaimed according to the actual occurrence of the business, so as to apply for the memory shard corresponding to other business, so as to improve the memory usage efficiency.

In one possible design, for a memory slice of any length, the amount of memory occupied by a memory slice of that length is managed in the form of a queue of unused memory blocks, a queue of exhausted memory blocks, and a queue of memory blocks in use. the memory address of the memory block;

The unused memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are free;

The exhausted memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are occupied;

The memory block queue in use includes memory addresses of memory blocks occupied by some of the memory segments in the memory segments of the length.

Through the above-mentioned memory management method, multiple function queues are used to manage the memory blocks occupied by memory fragments of a certain length, which can greatly reduce the delay when a business applies for memory. In addition, page table switching can be reduced, and the amount of computation required for looking up and switching page tables and the like can be reduced.

In a possible design, when a service request is received, allocating an idle memory segment of a corresponding length to the service request includes:

When the service request is received, from the free memory fragments of the length corresponding to the service request, the free memory fragment in the memory block with the smallest number of free memory fragments among the occupied memory blocks is preferentially allocated.

Through the above memory management method, the continuity of data writing in the memory can be ensured, and excessive memory fragmentation can be avoided.

In a possible design, for a memory slice of any length, a multi-level Bitmap lookup table is used to maintain the occupancy of the memory slice of the length.

Through the multi-level bitmap Bitmap lookup table, the actual occupancy of memory slices can be maintained in an orderly manner.

In a possible design, the method further includes:

For memory blocks of any size, detecting whether the proportion of memory blocks in an occupied state in the memory blocks of the size is less than a first preset threshold;

If the ratio is less than the first preset threshold, unload at least one memory block in an idle state among the memory blocks of the size.

In a possible design, the method further includes:

For memory blocks of any size, detecting whether the proportion of memory blocks in the occupied state in the memory blocks of the size is greater than a second preset threshold;

If the ratio is greater than the second preset threshold, load at least one memory block in an idle state from the system memory.

By determining whether to load or unload from the system memory according to the actual occupancy of memory blocks, flexible memory management can be achieved, and it can also meet the data volume requirements of services in different scenarios.

In a possible design, the method further includes:

For memory slices of any length, cache the memory addresses of a preset number of free memory slices of the length in the form of a double-pointer LIFO cache;

Correspondingly, when a service request is received, the memory address of the free memory slice is allocated from the double-pointer LIFO cache corresponding to the service request.

Through the above memory management method, the memory addresses of some memory segments are obtained in advance for allocation and use for business requests, which can reduce the time required for locating free memory segments and greatly reduce the delay of business requesting memory.

In a possible design, the method further includes:

For memory slices of any length, detect that the number of memory addresses in the double-pointer LIFO cache is lower than a preset threshold;

If the number of memory addresses is lower than the preset threshold, the memory addresses of a preset number of memory slices are acquired and stored in the double-pointer LIFO cache.

Through the above memory management method, the number of memory addresses in the cache can be guaranteed, which greatly reduces the delay of business application for memory.

Thirdly, an embodiment of the present invention also provides a data communication device, the data communication device includes a receiving device, a sending device, a system host, a memory management device, and a memory, wherein the memory management device is configured to monitor at least two memory blocks. state; when a memory block has been fragmented, the state of the memory block is an occupied state, and when a memory block has not been fragmented, the state of the memory block is an idle state; monitoring at least two The state of memory shards, the at least two memory shards include at least two shard lengths, and each memory shard is divided into memory blocks; if the number of free memory shards of the same length is less than the memory block application threshold, Then, according to the state of the at least two memory blocks, apply for a memory block in an idle state for division; if the number of free memory fragments of the same length is greater than the memory block release threshold, then release the free memory fragments of the same length; When a service request is received, an idle memory segment of a corresponding length is allocated to the service request; and data corresponding to the service request is stored according to the memory address of the allocated idle memory segment.

The data communication device provided by the present invention realizes the dynamic management of memory blocks by performing hierarchical management on memory blocks and memory slices, and flexibly applies for or releases memory blocks according to the actual occupancy of memory slices, thereby realizing the dynamic management of memory blocks and meeting business needs. The burstiness of memory resources enhances the sharing of memory resources, avoids resource waste, and improves memory usage efficiency.

In a possible design, the method further configures the memory management device to release the memory fragment of any length if no application for the memory fragment of the length is received within a preset time. The memory block occupied by the memory fragment of the specified length.

Through the above-mentioned data communication device, it is possible to judge whether the memory block needs to be reclaimed according to the actual occurrence of the business, so as to apply for the memory shard corresponding to other business, so as to improve the efficiency of memory usage.

In one possible design, for a memory slice of any length, the amount of memory occupied by a memory slice of that length is managed in the form of a queue of unused memory blocks, a queue of exhausted memory blocks, and a queue of memory blocks in use. the memory address of the memory block;

The unused memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are free;

The exhausted memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are occupied;

The memory block queue in use includes memory addresses of memory blocks occupied by some of the memory segments in the memory segments of the length.

Through the above data communication equipment, multiple function queues are used to manage the memory blocks occupied by a certain length of memory shards, which can greatly reduce the delay when a business applies for memory. In addition, page table switching can be reduced, and the amount of computation required for looking up and switching page tables and the like can be reduced.

In a possible design, the memory management device is configured to, when receiving the service request, preferentially allocate the least number of free memory fragments in the occupied memory block from the free memory fragments of the length corresponding to the service request. free memory slices in the memory block of .

Through the above data communication device, the continuity of data writing in the memory can be ensured, and excessive memory fragmentation can be avoided.

In a possible design, for a memory slice of any length, a multi-level Bitmap lookup table is used to maintain the occupancy of the memory slice of the length.

Through the multi-level bitmap Bitmap lookup table, the actual occupancy of memory slices can be maintained in an orderly manner.

In a possible design, the memory management device is configured to, for a memory block of any size, detect whether the proportion of the memory blocks in the occupied state in the memory blocks of the size is less than a first preset threshold; if the If the ratio is smaller than the first preset threshold, at least one memory block in an idle state among the memory blocks of the size is unloaded.

In a possible design, the memory management device is configured to, for a memory block of any size, detect whether the proportion of the memory blocks in the occupied state in the memory blocks of the size is greater than a second preset threshold; if the If the ratio is greater than the second preset threshold, at least one memory block in an idle state is loaded from the system memory.

By determining whether to load or unload from the system memory according to the actual occupancy of memory blocks, flexible memory management can be achieved, and it can also meet the data volume requirements of services in different scenarios.

In a possible design, the memory management device is configured to, for a memory slice of any length, cache the memory of a preset number of free memory slices of the length in the form of a double-pointer LIFO cache. Correspondingly, when a service request is received, the memory address of the free memory slice is allocated from the double-pointer LIFO cache corresponding to the service request.

Through the above data communication device, the memory addresses of some memory segments are obtained in advance for allocation and use for business requests, which can reduce the time required for locating free memory segments and greatly reduce the delay of business requesting memory.

In a possible design, the memory management device is configured to detect that the number of memory addresses in the double-pointer LIFO cache is lower than a preset threshold for memory slices of any length;

If the number of memory addresses is lower than the preset threshold, the memory addresses of a preset number of memory slices are acquired and stored in the double-pointer LIFO cache.

Through the above-mentioned data communication device, the number of memory addresses in the cache can be guaranteed, which greatly reduces the delay of business application for memory.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

FIG. 1 is a schematic diagram of a hardware structure of a data communication device provided by an embodiment of the present invention.

FIG. 2 is a schematic structural diagram of a memory management apparatus provided by an embodiment of the present invention.

FIG. 3 is a schematic diagram of a queue maintained by a memory block management module according to an embodiment of the present invention.

FIG. 4 is a schematic diagram of a three-level Bitmap lookup table provided by an embodiment of the present invention.

FIG. 5 is a schematic diagram of a configuration manner of a memory block queue and a memory fragmentation queue provided by an embodiment of the present invention.

FIG. 6 is a flowchart of a memory management method provided by an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

An embodiment of the present invention provides a memory management device, which is especially suitable for data communication equipment. In order to make it easier for readers to understand the technical solutions of the embodiments of the present invention, the present invention will hereinafter describe the hardware structure of the above-mentioned data communication device. Briefly explain.

FIG. 1 is a schematic diagram of the hardware structure of the above data communication device. The data communication device may at least include a receiving device 110 , a sending device 120 , a system host 130 , a memory management device 140 and a memory 150 . The receiving device 110 is used to receive data packets, the sending device 120 is used to send data packets, the system host 130 is used to perform specific processing on the received data packets, and the memory management device 140 is used to During the processing, memory is allocated for the data corresponding to the service request according to the service request generated during the processing.

Please refer to FIG. 2 , which shows a schematic structural diagram of a memory management apparatus provided by an embodiment of the present invention. As shown in FIG. 2 , the apparatus may include, but is not limited to, a memory block status monitoring module 210 , a memory fragmentation monitoring module 220 , a memory fragmentation management module 230 and a data storage module 240 . The following describes the functions of each module:

The memory block status monitoring module 210 .

The memory block state monitoring module 210 is used to monitor the states of at least two memory blocks; when a memory block has been fragmented, the state of the memory block is an occupied state, and when a memory block has not been fragmented, Then the state of the memory block is an idle state.

The memory 150 can be divided into multiple memory blocks, and the sizes of the multiple memory blocks can be set by technicians according to the requirements of the data communication service. In addition, the sizes of the multiple memory blocks can be the same or different; In practical applications, the memory management configuration module 210 can also divide the memory into a whole memory block. The present invention does not specifically limit the memory block division methods such as the number of memory blocks and the size of the memory blocks.

For data communication equipment, it has the ability to process a variety of different services. Therefore, in order to meet the needs of various services, each service can occupy a certain number of memory blocks in advance, and each service can occupy at least one of the same size. It can also occupy at least two memory blocks of different sizes, which are not specifically limited here. In order to avoid waste of memory resources, the occupied memory block can be further divided into at least one memory shard, so that memory application and allocation can be performed in units of memory shards for business requests. It should be noted here that the size of each memory slice in each memory block can also be set by technicians according to the requirements of data communication services, and the number of memory slices in each memory block can be the same or different, and , the size of each memory slice in each memory block may be the same or different, which is not specifically limited in the present invention.

It should be noted that, when the memory 150 is divided into memory blocks, only a part of the memory of the memory 150 may be divided, and the remaining part of the memory may not be divided temporarily for subsequent flexible use. Further, in order to cope with the burst of service requests, the number of enabled memory blocks can be flexibly adjusted according to the occupancy of the memory blocks. Specifically, the memory block status monitoring module can be used in any of the following implementation processes:

In the first implementation process, the memory block status monitoring module is used for memory blocks of any size to detect whether the proportion of the memory blocks in the occupied state in the memory blocks of the size is smaller than the first preset threshold; if the If the ratio is smaller than the first preset threshold, at least one memory block in an idle state among the memory blocks of the size is unloaded.

For this implementation process, if the proportion of the memory blocks in the occupied state among the memory blocks of the size detected is smaller than the first preset threshold, it means that the current amount of data required by the service is small, and the currently occupied memory blocks are too many , in order to avoid wasting memory resources, the memory block occupied by the service can be unloaded, that is, the memory block is released back to the memory for other services to load the memory block.

The second implementation process, the memory block status monitoring module is used for memory blocks of any size to detect whether the proportion of the memory blocks in the occupied state in the memory blocks of the size is greater than the second preset threshold; if the If the ratio is greater than the second preset threshold, at least one memory block in an idle state is loaded from the system memory.

For this implementation process, if it is detected that the proportion of memory blocks in the occupied state in the memory blocks of the size is greater than the second preset threshold, it means that the current remaining amount of memory blocks of the size is insufficient, which may not be enough to deal with the problem of the service. Therefore, at least one free memory block can be loaded from the system memory to ensure that business processing is not affected.

It should be noted that, in addition to judging whether to unload or load a memory block according to the above ratio, judgment may also be made according to the number of memory blocks in an occupied state, etc., which is not limited in this embodiment of the present invention.

For the management of memory blocks, it can be carried out in the form of queues. Each queue is used to store the addresses of memory blocks. Each queue can be used to manage memory blocks of the same size, and the number of addresses stored in each queue can be the same or different. , which is not specifically limited in this embodiment of the present invention. It should be noted that the queue exists in the form of a linked list or other data structures, which may at least include the following information: pointers to the head and tail of the queue, the number of memory blocks, and the size of the memory blocks.

In summary, the above-mentioned loading and unloading of memory blocks according to the actual occupancy of the memory blocks enables the memory blocks to be dynamically adjusted according to the actual needs of the business, thereby enhancing the sharing of memory resources and improving the efficiency of memory use.

Memory fragmentation monitoring module 220

The memory fragmentation monitoring module 220 is used to monitor the status of at least two memory fragments, the at least two memory fragments include at least two fragment lengths, and each memory fragment is divided into memory blocks; If the number of free memory fragments is less than the memory block application threshold, then according to the status of the at least two memory blocks, apply for the memory block in the free state to be divided; if the number of free memory fragments of the same length is greater than the memory block release threshold, then Free memory segments of the same length are released.

Among them, the state of the memory segment includes the occupied state and the idle state. When the memory segment has been allocated, the memory segment is in the occupied state. When the memory segment has been released or has not been allocated, the memory segment is in the occupied state. The slice is idle. However, if the number of free memory shards of the same length is less than the memory block application threshold, it means that the memory shards of this length may not be enough to meet the business requirements, you can apply for at least one memory block from the memory blocks in the idle state, and assign The applied memory block is divided into at least one memory shard for business application needs. If a memory block in an idle state is allocated and divided, the state of the memory block is converted to the occupied state. However, if the number of free memory fragments of the same length is greater than the memory block release threshold, it means that the amount of business data corresponding to the memory fragments of this length is limited. In order to avoid wasting memory resources, the free memory fragments of the same length can be released. piece. When releasing, the memory addresses of the memory slices can be aggregated to obtain the address of the entire memory block, and then the address of the memory block is released.

Further, the memory fragmentation monitoring module is used for the memory fragmentation of any length, if the allocation of the memory fragmentation of the length is not received within the preset time, then the memory fragmentation of the length is released. The block of memory occupied by the slice. If no application is received for a memory slice of this length within the preset time, it means that there is no demand for the service, and in order to avoid wasting memory resources, the memory blocks occupied by the memory slice of this length can be released.

In order to reduce memory fragmentation and increase the efficiency of memory use, for a memory fragment of any length, multiple queues with different functions can be used to manage the memory blocks occupied by the memory fragment of this length, see Figure 3, and the details are as follows: The memory fragmentation monitoring module is used for memory fragmentation of any length, in the form of unused memory block queues, exhausted memory block queues, and in-use memory block queues to manage the occupied memory fragments of the lengths. The memory address of the memory block. Wherein, the unused memory block queue includes memory addresses of memory blocks in which all memory fragments of the length are free. The exhausted memory block queue includes memory addresses of memory blocks in which all memory fragments of the length are occupied. The memory block queue in use includes memory addresses of memory blocks occupied by some of the memory segments in the memory segments of the length.

It should be noted that a memory block occupied by a memory fragment can have three states: unused state, that is, the memory fragments divided by the memory block are all in an idle state; used state, that is, the memory block Some of the divided memory shards are in the idle state, and some have been allocated; in the exhausted state, that is, all the memory shards divided by the memory block have been allocated. With the operation of the device, the state of the memory fragment will change, which will cause the state of the memory block to change at any time. For example, when a memory fragment in an unused memory block is allocated, the unused memory The memory block immediately becomes a memory block in use. In order to truly reflect the changes in memory, the above queue needs to be dynamically maintained. The dynamic maintenance process can include the following situations (a) to (d):

a) When all the memory fragments in the first specified memory block in the memory block queue are free, the memory fragment monitoring module deletes the first specified memory block from the memory block queue in use and adds it to the unused memory Tail of the block queue.

b) When the memory block queue in use does not include any memory blocks, the memory fragmentation monitoring module obtains at least one free memory block from the head of the unused memory block queue, and adds the at least one free memory block to the to the end of the queue of memory blocks in use.

c) When at least one memory segment of the second specified memory block in the exhausted memory block queue becomes free, the memory segment monitoring module deletes the second specified memory block from the exhausted memory block queue, It is added to the head of the memory block queue in use, so that the memory fragments in the memory block can be allocated preferentially, so as to reduce the memory fragmentation and increase the efficiency of memory usage.

d) When the memory segments of the third specified memory block in the memory block queue are all allocated, the memory segment monitoring module deletes the third specified memory block from the memory block queue in use and adds it to the consumed memory block. The tail of the memory block queue is exhausted.

In summary, by managing the memory blocks occupied by memory fragments of a certain length in the form of a variety of different functional queues, the memory fragments in the memory blocks in use can be allocated preferentially, thereby reducing memory fragmentation and increasing memory usage efficiency. , improve the flexibility of memory allocation and avoid the negative impact of frequent page table switching,

Further, in order to improve the positioning speed of memory shards during subsequent memory application and reduce the delay of business application for memory, the present invention provides a maintenance method for the occupancy of memory shards, that is, the memory shard monitoring. The module is used to maintain the occupancy status of the memory segment of any length by adopting a multi-level Bitmap lookup table for memory segments of any length.

For example, for a memory slice of one length, a multi-level bitmap Bitmap lookup table can be configured, and the multi-level bitmap Bitmap lookup table includes: a first-level Bitmap scheduling table and a third-level Bitmap look-up table. The third-level Bitmap lookup table The look-up table includes: a first-level Bitmap look-up table, a second-level Bitmap look-up table and a third-level Bitmap look-up table.

Among them, the first-level Bitmap scheduling table is used to store the unused memory block queue, the exhausted memory block queue and the mapping relationship between the memory blocks in the memory block queue in use and the third-level Bitmap lookup table, the first The first-level Bitmap look-up table is a one-dimensional storage structure, including multiple first-level storage units, each first-level storage unit corresponds to a row of the second-level Bitmap look-up table, and the second-level Bitmap look-up table is a two-dimensional storage structure, including multiple rows and multiple secondary storage units, each secondary storage unit corresponds to a row of the third-level Bitmap look-up table, the third-level Bitmap look-up table is a two-dimensional storage structure, including multiple third-level storage units, each The third-level storage unit corresponds to one of the memory slices. FIG. 4 is a schematic diagram of the above-mentioned three-level Bitmap lookup table. The first-level storage unit, the second-level storage unit and the third-level storage unit all correspond to a preset value , the preset value is used to reflect the occupancy status of the memory corresponding to the first-level storage unit, the second-level storage unit, and the third-level storage unit. For example, the preset value can be 0 or 1, where 0 is used to represent the memory Not occupied or not fully occupied, 1 is used to indicate occupied, it should be noted that the above preset value is only an example, in practical application, the preset value can be any value, and the preset value There may be more than two, which is not specifically limited in the present invention. In addition, in practical applications, the multi-level bitmap Bitmap lookup table may only include the Bitmap lookup table, but not the Bitmap scheduling table, or may include four levels, two The Bitmap lookup table of any series of levels is not specifically limited in the present invention, and since the lookup methods of the Bitmap lookup tables of different levels are very similar, the present invention will use the three-level Bitmap lookup table as An example is used to illustrate, and the search methods of the Bitmap lookup table of other series will not be described again.

In order to facilitate the understanding of the above-mentioned method for locating free memory slices in the memory block queue in use by using the multi-level Bitmap lookup table, the present invention will briefly illustrate the method.

For example, if you need to find the first free memory slice in the memory block queue in use, you can first look up the first-level Bitmap scheduling table to obtain the third-level Bitmap lookup table corresponding to the memory block at the head of the memory block queue in use, and then The first-level storage unit with a default value of 0 can be searched in order in the first-level Bitmap lookup table (the default value of 0 means that the memory is not occupied or not fully occupied), assuming the first preset value queried. The number of the first-level storage unit that is 0 is a, then the second-level storage unit with a preset value of 0 can be searched in order in the a-th row of the second-level Bitmap lookup table, assuming that the first preset value queried. The number of the second-level storage unit that is 0 is b, then the memory management apparatus 200 can search for the third-level storage unit with a preset value of 0 in order in the abth row of the third-level Bitmap lookup table, assuming that the queried first The number of the three-level storage unit with the default value of 0 is ijkl0124, then the memory slice numbered ijkl0124 is the first free memory slice in the memory block queue in use. It should be noted that in the above example, the The numbers are only exemplary and do not limit the invention.

When using queues to manage memory blocks and memory fragments, the logical relationship between memory block queues and memory fragmentation queues can be seen in Figure 5. As shown in Figure 5, multiple memory fragmentation queues can correspond to one memory In the block queue, the memory segments managed by each memory segment queue are divided by the memory blocks managed by the memory block queue. In an actual scenario, the memory 150 can be managed by multiple memory block queues.

Memory slice management module 230

The memory fragmentation management module 230 is configured to allocate free memory fragments of a corresponding length to the service request when receiving a service request.

In order to reduce memory fragmentation, when allocating memory fragments, the memory fragment in the memory block at the head of the memory block queue can be allocated first. Since the corresponding relationship between logical addresses and physical addresses is stored in the page table, when allocating memory slices, it is necessary to use the page table to obtain the mapping relationship between the memory addresses of the memory slices and the memory blocks, and when allocating memory When fragmenting, the memory fragment in the memory block in the head of the memory block queue is always allocated, which makes it possible to obtain the above mapping relationship without switching the page table. Therefore, this method can effectively suppress the frequent switching of the page table. negative impact.

In a possible design, the memory fragmentation management module 230 is configured to, when receiving a service request, preferentially allocate the number of free memory fragments in the occupied memory block from the free memory fragments of the length corresponding to the service request Free memory slices in the fewest memory blocks.

The inventor realized that, in addition to prioritizing the allocation of memory segments in the memory block in use, it is also possible to prioritize the allocation of memory segments in the memory block in use with a smaller number of free memory segments, which can also reduce memory usage. Fragmentation has the effect of increasing the efficiency of memory usage. Therefore, in an embodiment of the present invention, for each length of memory slices, the number of free memory slices in each memory block included in the memory block queue can be detected every predetermined time, and the number of free memory slices can be determined according to The number of free memory shards is in ascending order, and the queue of memory blocks in use is sorted, so that the memory shards in the memory block with the least number of free memory shards are allocated first. The time can be set by a skilled person, which is not specifically limited in the present invention.

In a possible design, the memory fragmentation management module is configured to, for memory fragments of any length, cache a preset number of free memory fragments of the length in the form of a double-pointer LIFO cache. The memory address of the slice, when a service request is received, the memory address of the free memory slice is allocated from the double-pointer LIFO cache corresponding to the service request.

In order to further reduce the delay of business applying for memory, a pair of pointer LIFO caches can be provided for each length of memory shard to cache a certain number of memory addresses, so that when a business applies for memory shards, Direct allocation from this cache may be preferred. When the maintenance form of the multi-level bitmap Bitmap lookup table is adopted, the above steps of querying the multi-level bitmap Bitmap lookup table can be avoided, and the memory can be directly allocated from the cache, thereby further reducing the delay of business application memory and reducing the memory Manage the amount of computation required.

For the above scenario in which the memory addresses of a preset number of free memory slices of the described length are cached in the form of a double-pointer LIFO cache, when a memory slice usage request is received, the stack top pointer can be stored in the corresponding memory The memory addresses of the to-be-allocated memory slices in the unit are allocated first. After receiving the memory release request, it can also store the memory address of the memory segment to be released into the cache, that is, write the memory address of the memory segment to be released into the storage unit corresponding to the stack top pointer, so that the It can be allocated preferentially, which can avoid frequent switching of the page table when the business applies for memory.

In a possible design, the memory fragmentation management module 230 is configured to, for memory fragments of any length, detect that the number of memory addresses in the double-pointer LIFO cache is lower than a preset threshold; if If the number of memory addresses is lower than the preset threshold, the memory addresses of a preset number of memory slices are acquired and stored in the double-pointer LIFO cache.

Wherein, if the number of stored memory addresses is lower than the preset threshold, indicating that the to-be-allocated memory slices are insufficient, the memory addresses of the preset number of memory slices can be obtained from the memory slices of this length. It should be noted that, the above-mentioned prefetch threshold and preset number can be set by a technician, which is not specifically limited in the present invention. For example, taking the above-mentioned memory block queue as an example, the memory slice management module can search for the free memory in the memory block in use in the multi-level Bitmap lookup table hierarchically according to the arrangement order of the memory blocks in the memory block queue in use. Fragment, the memory block in use is a memory block occupied by a part of the memory fragment, obtain a preset number of free memory fragments in the search order, and obtain the memory address of the preset number of free memory fragments into the cache .

In order to help readers understand the above technical process, the above acquisition process is described below based on the structure of the multi-level bitmap Bitmap lookup table. For example, by comparing the number of memory segments to be allocated stored in the cache with the prefetch threshold, it is found that 10 free memory segments need to be acquired, then the multi-level bitmap Bitmap lookup table can be queried. Specifically, First, look up the first-level Bitmap scheduling table to obtain the third-level Bitmap lookup table corresponding to the first memory block at the head of the memory block queue being used. Level storage unit (the default value of 0 means that the memory is not occupied or not fully occupied). Assuming that the number of the first level storage unit with a default value of 0 is queried, it can be stored in the second level Bitmap In the a-th row of the lookup table, the secondary storage unit with the preset value of 0 is searched in order. Assuming that the number of the first secondary storage unit with the preset value of 0 is b, you can use the third-level Bitmap In the abth row of the lookup table, the tertiary storage units with the preset value of 0 are searched in order. If 10 third-level storage units with the preset value of 0 are found in the abth row of the third-level Bitmap lookup table, then Prefetch the memory addresses of the memory slices corresponding to the above-mentioned 10 third-level storage units with a default value of 0 into the above cache, if the default value found in the abth row of the third-level Bitmap lookup table is 0 If there are less than 10 third-level storage units, then it returns to continue to search for the second-level storage unit with a preset value of 0 in the a-th row of the second-level Bitmap lookup table, assuming that the second preset value queried by the data storage module 230 The number of the second-level storage unit with a value of 0 is c, then the third-level storage unit with a preset value of 0 can be searched in order in the ac-th row of the third-level Bitmap lookup table, and the above query process can be executed cyclically until it is 10 memory addresses will be obtained. Of course, if enough memory fragments are not found after returning to the second-level Bitmap lookup table, it can continue to return to the first-level Bitmap lookup table or even the first-level Bitmap scheduling table for query. It should be noted that the above-mentioned preset values and the numbers of the storage units at all levels are only exemplary, and do not limit the present invention.

It should be pointed out that after the above-mentioned memory addresses of the preset number of free memory slices are obtained into the cache, the above-mentioned preset number of free memory slices can be corresponding to the three-level storage unit in the multi-level bitmap Bitmap lookup table. The preset value of , is modified to indicate that the memory addresses of the preset number of free memory slices have been acquired into the above cache, so as to avoid errors in the next acquisition process.

And further, for a memory fragment of any length, it is detected whether the number of memory addresses in the double-pointer LIFO cache is higher than the recycling threshold; if the number of memory addresses is higher than the recycling threshold, then The two-pointer LIFO cache releases the memory addresses of the preset number of memory slices. The release may refer to reclaiming the memory fragment into the three-level Bitmap lookup table.

It should also be pointed out that the storage structure of the above cache can be various, for example, the above cache can be a LIFO cache, or a FIFO cache, etc., which is not specifically limited in the present invention. When the cache is a double-pointer LIFO cache, the cache can include a stack top pointer and a stack bottom pointer, and a preset number of free memory slices can be written into the storage corresponding to the stack bottom pointer of the double-pointer LIFO cache. in the unit.

data storage module 240

The data storage module 240 is configured to store the data corresponding to the service request according to the memory addresses of the free memory segments allocated by the memory segment management module. The data storage module 240 serves as a front end for processing service requests, and its specific storage process is not described here.

The device provided by the embodiment of the present invention implements dynamic management of memory blocks by performing hierarchical management on memory blocks and memory slices, and flexibly applies for or releases memory blocks according to the actual occupation of memory slices, thereby realizing dynamic management of memory blocks and meeting business needs. The burstiness of memory resources enhances the sharing of memory resources, avoids resource waste, and improves memory usage efficiency. In addition, the memory addresses of some to-be-allocated memory slices can be pre-obtained by means of caching, so that when a service applies for memory, the memory addresses of the memory slices in the cache are directly allocated to the service, without the need for idling The query of memory shards can reduce the delay of business application for memory.

FIG. 6 is a flowchart of a memory management method provided by an embodiment of the present invention. Referring to Figure 6, this embodiment specifically includes:

601. Monitor the states of at least two memory blocks; when a memory block has been fragmented, the state of the memory block is an occupied state, and when a memory block has not been fragmented, the state of the memory block is occupied. The state is idle.

602. Monitor the status of at least two memory segments, where the at least two memory segments include at least two segment lengths, and each memory segment is divided into memory blocks.

603. If the number of free memory fragments of the same length is less than the memory block application threshold, then according to the states of the at least two memory blocks, apply for a memory block in an idle state for division; if the number of free memory fragments of the same length is greater than If the memory block release threshold is set, then free memory fragments of the same length are released.

It should be noted that the execution sequence of the above steps 601 to 603 may be parallel or any execution sequence, and the monitoring periods may be the same or different, which is not specifically limited in this embodiment of the present invention.

604. When a service request is received, allocate an idle memory segment of a corresponding length to the service request.

605. Store data corresponding to the service request according to the allocated memory addresses of the free memory slices.

In one possible design, the method further includes:

For a memory fragment of any length, if no application for the memory fragment of the length is received within a preset time, the memory block occupied by the memory fragment of the length is released.

In one possible design, for a memory slice of any length, the memory occupied by a memory slice of that length is managed in the form of a queue of unused memory blocks, a queue of exhausted memory blocks, and a queue of memory blocks in use the memory address of the block;

The unused memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are free;

The exhausted memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are occupied;

The memory block queue in use includes memory addresses of memory blocks occupied by some of the memory segments in the memory segments of the length.

In a possible design, when a service request is received, allocating an idle memory segment of a corresponding length to the service request includes:

When the service request is received, from the free memory fragments of the length corresponding to the service request, the free memory fragment in the memory block with the smallest number of free memory fragments among the occupied memory blocks is preferentially allocated.

In a possible design, for a memory slice of any length, a multi-level Bitmap lookup table is used to maintain the occupancy of the memory slice of the length.

In one possible design, the method further includes:

For memory blocks of any size, detecting whether the proportion of memory blocks in an occupied state in the memory blocks of the size is less than a first preset threshold;

If the ratio is less than the first preset threshold, unload at least one memory block in an idle state among the memory blocks of the size.

In one possible design, the method further includes:

For memory blocks of any size, detecting whether the proportion of memory blocks in the occupied state in the memory blocks of the size is greater than a second preset threshold;

If the ratio is greater than the second preset threshold, load at least one memory block in an idle state from the system memory.

In one possible design, the method further includes:

For memory slices of any length, cache the memory addresses of a preset number of free memory slices of the length in the form of a double-pointer LIFO cache;

Correspondingly, when a service request is received, the memory address of the free memory slice is allocated from the double-pointer LIFO cache corresponding to the service request.

In one possible design, the method further includes:

For memory slices of any length, detect that the number of memory addresses in the double-pointer LIFO cache is lower than a preset threshold;

If the number of memory addresses is lower than the preset threshold, the memory addresses of a preset number of memory slices are acquired and stored in the double-pointer LIFO cache.

It should be noted that, the specific implementation process of the foregoing method embodiment is the same as that described in the apparatus embodiment, and is not repeated here.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, for the specific process of the method described above, reference may be made to the corresponding process in the apparatus embodiment, which is not repeated here.

Those of ordinary skill in the art can understand that all or part of the steps of implementing the above embodiments can be completed by hardware, or can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The storage medium mentioned may be a read-only memory, a magnetic disk or an optical disk, etc.

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (16)

1. A memory management device, comprising: The memory block status monitoring module is used to monitor the status of at least two memory blocks; when a memory block has been fragmented, the status of the memory block is occupied, and when a memory block has not been fragmented, Then the state of the memory block is an idle state; a memory fragmentation monitoring module, configured to monitor the status of at least two memory fragments, the at least two memory fragments include at least two fragment lengths, and each memory fragment is divided into memory blocks; The memory fragmentation monitoring module is further configured to, if the number of free memory fragments of the same length is less than the memory block application threshold, apply for a memory block in an idle state for division according to the states of the at least two memory blocks; if If the number of free memory fragments of the same length is greater than the memory block release threshold, the free memory fragments of the same length are released; The memory fragmentation management module is configured to preferentially allocate the free memory fragment in the memory block with the least number of free memory fragments among the occupied memory blocks from the free memory fragments of the corresponding length of the service request when receiving a service request ; A data storage module, configured to store the data corresponding to the service request according to the memory addresses of the free memory segments allocated by the memory segment management module. 2. The device according to claim 1, wherein the memory fragmentation monitoring module is used for: For a memory fragment of any length, if no application for the memory fragment of the length is received within a preset time, the memory block occupied by the memory fragment of the length is released. 3. device according to claim 1, is characterized in that, described memory fragmentation monitoring module is used for the memory fragmentation of any kind of length, with unused memory block queue, exhausted memory block queue and being used. The memory address of the memory block occupied by the memory segment of the described length is managed in the form of a memory block queue; The unused memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are free; The exhausted memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are occupied; The memory block queue in use includes memory addresses of memory blocks occupied by some of the memory segments in the memory segments of the length. 4. device according to claim 1, is characterized in that, described memory fragmentation monitoring module is used for the memory fragmentation of any kind of length, adopts the mode of multi-level bitmap Bitmap lookup table to maintain the memory of described length Shard occupancy. 5. The device according to claim 1, wherein the memory block status monitoring module is further used for: For memory blocks of any size, detecting whether the proportion of memory blocks in an occupied state in the memory blocks of the size is less than a first preset threshold; If the ratio is less than the first preset threshold, unload at least one memory block in an idle state among the memory blocks of the size. 6. The device according to claim 1, wherein the memory block status monitoring module is further used for: For memory blocks of any size, detecting whether the proportion of memory blocks in the occupied state in the memory blocks of the size is greater than a second preset threshold; If the ratio is greater than the second preset threshold, load at least one memory block in an idle state from the system memory. 7. device according to claim 1, is characterized in that, described memory fragmentation management module, for the memory fragmentation of any kind of length, in the form of double pointer LIFO cache, cache preset quantity When a service request is received, the memory address of the free memory fragment is allocated from the double-pointer LIFO cache that responds to the service request. 8. device according to claim 7, is characterized in that, described memory fragmentation management module, is used for the memory fragmentation of any kind of length, detects the memory address quantity in described double pointer LIFO cache memory below the preset threshold; If the number of memory addresses is lower than the preset threshold, the memory addresses of a preset number of memory slices are acquired and stored in the double-pointer LIFO cache. 9. A memory management method, comprising: Monitor the status of at least two memory blocks; when a memory block has been fragmented, the status of the memory block is occupied, and when a memory block has not been fragmented, the status of the memory block is idle state; Monitoring the status of at least two memory fragments, the at least two memory fragments include at least two fragment lengths, and each memory fragment is divided into memory blocks; If the number of free memory fragments of the same length is less than the memory block application threshold, then according to the states of the at least two memory blocks, apply for the memory blocks in the free state for division; if the number of free memory fragments of the same length is greater than the memory block If the threshold is released, the free memory fragments of the same length are released; When receiving a service request, from the free memory fragments of the corresponding length of the service request, preferentially allocate the free memory fragment in the memory block with the least number of free memory fragments among the occupied memory blocks; The data corresponding to the service request is stored according to the allocated memory addresses of the free memory segments. 10. The method according to claim 9, wherein the method further comprises: For a memory fragment of any length, if no application for the memory fragment of the length is received within a preset time, the memory block occupied by the memory fragment of the length is released. 11. The method of claim 9, wherein, for a memory slice of any length, the length is managed in the form of a queue of unused memory blocks, a queue of exhausted memory blocks, and a queue of memory blocks in use The memory address of the memory block occupied by the memory slice; The unused memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are free; The exhausted memory block queue includes the memory addresses of the memory blocks in which all memory fragments of the length are occupied; The memory block queue in use includes memory addresses of memory blocks occupied by some of the memory segments in the memory segments of the length. 12 . The method according to claim 9 , wherein, for a memory fragment of any length, a multi-level Bitmap lookup table is used to maintain the occupancy of the memory fragment of the length. 13 . 13. The method of claim 9, wherein the method further comprises: For memory blocks of any size, detecting whether the proportion of memory blocks in an occupied state in the memory blocks of the size is less than a first preset threshold; If the ratio is less than the first preset threshold, unload at least one memory block in an idle state among the memory blocks of the size. 14. The method of claim 9, wherein the method further comprises: For memory blocks of any size, detecting whether the proportion of memory blocks in the occupied state in the memory blocks of the size is greater than a second preset threshold; If the ratio is greater than the second preset threshold, load at least one memory block in an idle state from the system memory. 15. The method of claim 9, wherein the method further comprises: For memory slices of any length, cache the memory addresses of a preset number of free memory slices of the length in the form of a double-pointer LIFO cache; Correspondingly, when a service request is received, the memory address of the free memory slice is allocated from the double-pointer LIFO cache corresponding to the service request. 16. The method of claim 15, wherein the method further comprises: For memory slices of any length, detect that the number of memory addresses in the double-pointer LIFO cache is lower than a preset threshold; If the number of memory addresses is lower than the preset threshold, the memory addresses of a preset number of memory slices are acquired and stored in the double-pointer LIFO cache.

Images