JP2000047933A - Bidirectional memory securing management system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the fragmentation of a memory by searching a free area in the direction from a high address to a low address in a continuous memory space concerning a request for securing a size being equal to or larger than the one which is decided as against a memory securing request and securing it. SOLUTION: When the size requesting securing is equal to or larger than the size(St) which is decided against the memory securing request from a process in an information processor, a free memory block queue managed by a large free memory block managing part 51 is retrieved. When the request size to be secured is smaller than St, the free memory block queue managed by a small free memory block managing part 52 is retrieved. When the free memory block whose size is the same as the requested one is detected, it is adopted. Unless the equal sized one exists, the free memory block which is larger and also is provided with the size being the nearest to the requested size is adapted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a dynamic memory management function for storing a required amount of a work area on a main memory (memory) while a program (system-provided program or user program) operating on an information processing apparatus is being executed. The present invention relates to a memory management method of the dynamic memory management function in a process of requesting and dynamically securing the secured memory and requesting and releasing the secured memory when the secured memory becomes unnecessary.

[0002]

2. Description of the Related Art Conventionally, for a memory requested by a process (execution unit of a program operating on an information processing apparatus), a dynamic memory management function for managing the memory is provided by a continuous memory consisting of a memory securing free area. When the memory requested to be secured from the process is cut out from the space, only a unidirectional search such as searching for a free area from a low address to a high address in the continuous memory space is performed regardless of the size of the requested memory. Was.

The memory release processing, which is a part of the dynamic memory management function, is performed by a single free memory block for managing the free areas in one direction or two directions by linking the start addresses of the free areas with pointers. A free memory block (a memory block is a continuous area of the memory) which is a free area is registered in the queue, and when the free memory block is registered, if free areas before and after the free memory block are free and free, each other is registered. It has been common to create larger free memory blocks by concatenation.

In the case of a memory securing process, if there is a free memory block matching or larger than the size of the requested memory block in the free memory block queue, a required portion is cut out from the free memory block and requested. Given to the process, the rest of the area after clipping was also re-registered in the free memory block.

For this reason, if a large memory block is released and a part of the free large memory block is successively allocated to a plurality of small memory block reservation requests generated thereafter, In the released large memory block, an area where a small divided memory block is secured and a free area are mixed, so that a so-called memory fragmentation state occurs.

[0006] Therefore, if a request for securing a memory having the same size as the large memory is made next, the request cannot be fulfilled. Therefore, it is necessary to secure another new area, and the virtual memory space used by the process is enlarged. There was a problem of doing.

For example, in the case of the above-described conventional memory securing processing, a large-sized memory block of several megabytes or more is randomly and frequently secured and released from a small-sized memory block of several bytes to several hundred bytes. When executing a process (for example, a process that handles image data), the area allocated in units of several megabytes is released, connected to the free memory block queue, and then a request for allocation in units of several bytes, and then several hundred bytes If a request for securing a unit occurs randomly and the request size is cut out one after another from the free memory block of several megabytes, a large free memory block is subdivided, and then a new request of several megabytes is generated. The free memory block queue has a free Will be that the re-block does not exist, it is necessary to ensure the new address space of the newly several megabytes.

[0008] As a result, the memory size of the process gradually increases, and a problem occurs that the memory size cannot be secured after reaching the limit of the process size allowed for one process.

[0009]

Conventionally, both large and small memory requests are always in the same direction (such as the direction from the lower address to the higher address of the memory).
Assignment from was made. FIG. 8 shows the allocation status, in which allocation is performed from the low address direction to the high address direction, and an unused area exists in the high address portion.

[0010]

According to the bidirectional memory allocation management system of the present invention, a size smaller than a predetermined size (hereinafter referred to as St) is allocated in response to a memory allocation request from a process in an information processing apparatus. For requests, a free area is searched for and secured in the direction from a low address to a high address in a continuous memory space composed of free areas for securing memory. For a secure request of a size equal to or larger than St, Searching for and securing a free area from the high address to the low address direction is provided.

Further, in the bidirectional memory allocation management system of the present invention, with respect to the value of St, history information relating to the frequency of use for each requested size is collected for the requested size from the process, and the collected information is provided. It is prepared.

Further, the bidirectional memory reservation management system of the present invention comprises notifying the value of St from a process and reserving a memory based on the notification.

Further, according to the bidirectional memory allocation management method of the present invention, when the value of St is not notified from the process,
It is provided that memory is secured by a default value of St.

Further, according to the bidirectional memory allocation management method of the present invention, the size of the initial value of a continuous memory space consisting of the free area for memory allocation is specified by a process, and the specified size is set to the specified size. Providing a commensurate size of the memory space.

Further, according to the bidirectional memory securing management method of the present invention, a plurality of continuous memory spaces consisting of the memory securing free areas are prepared and the plurality of continuous memory spaces are prepared according to a memory securing request size from a process. One is selected, and St defined in each of the continuous memory spaces is selected.
In response to a request for securing a size smaller than St, a free area is searched for and secured from a low address to a high address in a continuous memory space composed of free areas for securing memory, and a size equal to or larger than St is secured. The request includes searching for and securing a free area from the high address to the low address in the continuous memory space.

[0016]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a first embodiment of a bidirectional memory allocation management system according to the present invention.

In this embodiment, a process 1, a memory allocation / release request analysis unit 2, a memory allocation size analysis unit 3, a free memory block search unit 4, a large free memory block management unit 51, and a small free memory block management unit 52 A large memory block address space reservation unit 61, a small memory block address space reservation unit 62, a requested memory block return unit 7, a memory release size analysis unit 8, a free memory block registration unit 9, and a dynamic memory management unit 10. Is done.

The process 1 issues a request to secure a memory required to execute its own processing, and after the processing, issues a request to release memory that is no longer needed.

The memory allocation / release request analysis unit 2 checks whether the request from the process 1 is a memory allocation request or a release request, and distributes the request to a process according to the request.

When the request from the process 1 is a memory reservation request, the memory reservation size analysis unit 3 analyzes the reservation size, rounds up to an appropriate size depending on the request size (such as a multiple of 4), and searches for a free memory block 4. Call.

The free memory block search unit 4 searches whether there is a free memory block of a size suitable for the allocation request size passed from the memory allocation size analysis unit 3. If the size requested to be secured is equal to or larger than St, a free memory block queue managed by the large free memory block management unit 51 is searched. When the required size to be secured is smaller than St, the free memory block queue managed by the small free memory block management unit 52 is searched.

If a free memory block equal to the requested size is detected, it is used. If there is no equal size, a larger free memory block having the size closest to the requested size is used. In this case, the required memory size is cut out from the free memory block and left. The size area is re-registered in the free memory block queue and returned.

If a free memory block of the requested size is not found, it is necessary to secure a new area. Therefore, if the requested size is equal to or larger than St, a new area is provided to the large memory block address space area securing unit 61. An area reservation request is made. If the area is smaller, a new area reservation request is made to the small memory block address space area reservation section 62.

The large free memory block management unit 51 is St
A processing unit that connects and manages free memory blocks that have been released with a size equal to or larger than the free memory block queue, and checks whether blocks before and after the free memory blocks registered in the free memory block queue are free. If it is free, the memory blocks are linked.

The small free memory block management unit 52 operates as St
This is a processing unit that connects and manages free memory blocks that have been freed to smaller sizes into the free memory block queue, and checks whether the blocks before and after the free memory block registered in the free memory block queue are free. For example, memory blocks are connected.

When the reserved memory size is equal to or larger than St, the large memory block address space reserving section 61 reserves a new memory from the memory space in the direction from the high address to the low address.

When the memory reservation size is smaller than St, the small memory block address space reservation unit 62 reserves a new memory from the memory space in the direction from the low address to the high address.

The request memory block return unit 7 returns to the process 1 the address of the memory block found in the free block or in which a new area has been secured.

When a memory release request is made from the process 1, the memory release size analysis unit 8 calculates the release request size and calls the free block registration unit 9.

The free memory block registration section 9 registers the free memory block in the large free memory block management section 51 or the small free memory block management section 52 according to the size of St.

The dynamic memory management unit 10 executes the above-described process 1 for executing a process for requesting memory reservation / release from the process 1.
Except for the entire configuration.

Next, the first embodiment (FIG. 1) of the present invention will be described in detail with reference to the flowchart of FIG.

Before the detailed description, a few supplementary explanations will be given. First, FIG. 4 schematically shows the structure of the process 1 to which the present invention is applied. According to the figure, a process 1 comprises a code section 31 composed of instruction codes, a static data section 32 for global / static data, and a dynamic data section 33 which is a memory area to be secured / released according to the present invention.
And a stack section 34 used to store caller information upon function call / return. Code section 31 and static data section 32
Is fixed and does not change from the start to the end of the execution of the process 1. Further, the dynamic memory management unit 10 of the present invention is executed by being statically or dynamically linked to the code section 31 as a library function.

When a memory reservation / release request is generated from the process 1, the memory reservation / release request analysis unit 2 in the dynamic memory management unit 10 is called and the request is accepted.
In the case of the first request from the process 1, the memory allocation / release request analysis unit 2
Then, an initialization process such as taking in the start address and end address of No. 3 is executed (step A1).

Next, the memory allocation / release request analysis unit 2
In (Step B1), it is determined whether the request from the process 1 is a memory reservation request or a memory release request. If the request is a reservation request, go to (Step C1), and if it is a release request, go to (Step D1). The following processing contents in the case of a memory securing request (step C1) will be described.

In the case of a memory reservation request, the memory reservation size analysis unit 3 determines that the memory reservation request size is high in order to speed up the address calculation processing if the memory reservation request size from the process 1 is not a multiple of 4 in (Step C1). Is rounded up to a multiple of 4, and the magnitude of the memory reservation request size Sx and St is compared.

If Sx is smaller than St (step E
Go to 1), and if Sx is equal to or greater than St, go to (Step E2).

The free memory block search unit 4 searches the free memory block queue for an unused area having a size equal to the requested size Sx in each of (Step E1) and (Step E2). Here, the structure of the free memory block queue is shown in FIG.

As shown in FIG. 5, a fixed length (size Mt) management information section is provided at the head of each free memory block. Note that the hatched area in the figure represents a memory block being used by the process 1, and although not shown, a management information section similarly exists.

The management information section has a pointer portion pointing bidirectionally to each free memory block and a portion indicating the size of each free memory block. So S
When there is a request to secure a memory block of size x, in the processing of the dynamic memory management unit 10, an area of size Sx + Mt is searched from the free memory block queue. Referring to FIG. 6, a memory block having a size of Sx + Mt is first secured from the free memory block queue, and then the forward and backward pointers of the memory block and the size Sx of the memory block are set in the management information section. Will be passed.

In the following description, although the management information portion having the size of Mt is omitted for the sake of explanation, the above-described processing is internally performed. If an area having a size equal to Sx cannot be detected in the free memory block queue, an area larger than Sx and closest to that size is searched.

In step (E1), the above processing is searched for a free memory block queue that manages the head area of the dynamic data section 33. In (Step E2), a search is made for the free memory block queue that manages the area behind the dynamic data section 33.

As a result of the search in (Step E1), when a free memory block corresponding to the requested size Sx can be searched, the procedure goes to (Step G1). As a result of the search in (Step E1), when a free memory block having a size corresponding to the requested size cannot be searched, the procedure goes to (Step H1).

In (Step E2), contrary to the above (Step E1), as a result of searching the free memory block queue for managing the rear side, if a free memory block corresponding to the requested size Sx can be searched, (Step G
Go to 2). As a result of the search in (Step E2), when a free memory block having a size corresponding to the requested size cannot be searched, the procedure goes to (Step H2).

This processing is performed when there is a memory block having a size corresponding to the requested size Sx in the free memory block queue for managing the head area (step G).
In 1), the small free memory block management unit 52 fetches a free block conforming to the size Sx, and after fetching, if there is any remaining portion, re-registers it in the free block queue and returns it.

In the processing when there is no memory block having a size corresponding to the requested size Sx in the free memory block queue managing the head area (step H1), the small memory block address space area securing section 62 Is the requested size Sx from the top of the unused area
Cut out the memory corresponding to.

This processing is performed when a memory block having a size corresponding to the required size Sx exists in the free memory block queue for managing the rear area (step G).
In 2), the large free memory block management unit 51 fetches a free block conforming to the size Sx, and after fetching, if there is any remaining portion, re-registers it in the free block queue and returns it.

In the processing when there is no memory block having a size corresponding to the requested size Sx in the free memory block queue for managing the rear area (step H2), the large memory block address space area securing section 61 Is the requested size Sx from the rear side of the unused area.
Cut out the memory corresponding to. In (Step I1), the above-mentioned management information is set. Then go to (Step J1).

Next, processing for a memory release request will be described.

If the request from the process 1 is a memory release request in (Step B1), (Step D
Going to 1), here the memory release size analysis unit 8 calculates the memory size Sx to be released, and the free block registration unit 9 compares the memory size Sx with the determined value St. When the release request size Sx is smaller than St, (Step F1) is called, and when Sx is larger than or equal to St, (Step F2) is called.

In (Step F1), the memory blocks released to the free memory block queue for managing the head area by the small free memory block management section 52 are re-registered in the ascending order of the addresses of the memory blocks on the queue. . If blocks before and after the address of the registered memory block are free, concatenate the blocks and combine them into one, and if the released memory block is continuous with the unused area, unused the memory block Reregister in the area section and go to (Step J1).

In (Step F2), the memory blocks released to the free memory block queue for managing the rear area by the large free memory block management unit 51 are re-registered in descending order of the addresses of the memory blocks on the queue. . If blocks before and after the address of the registered memory block are free, concatenate the blocks and combine them into one, and if the released memory block is continuous with the unused area, unused the memory block Reregister in the area section and go to (Step J1). The following is a common process for securing and releasing memory.

After performing the memory reservation request and the memory release request processing as described above, in (step J1), the request memory block returning unit 7 returns the address of the reserved memory block and the reservation result in the case of the memory reservation request. Is returned to process 1. In the case of a memory release request, the release result is returned to the process 1 and the process is terminated.

Referring to FIG. 7 to summarize the above description, an area smaller than St is allocated from the low address side, while an area having a size equal to or larger than the St value is allocated from the high address side. ing.

Next, a second embodiment (FIG. 3) of the present invention will be described in detail with reference to the drawings.

Referring to FIG. 3, according to a second embodiment of the present invention, a dynamic memory management unit 10 in the dynamic memory management unit 10 according to the first embodiment shown in FIG. 1 in that the memory reservation / release request processing unit 2 is a process request analysis unit 21 and that it has an St storage unit 11. Since the other components are the same as those of the first embodiment, the description will be focused on the process request analysis unit 21 and the St storage unit 11.

It is desirable that the value of St be determined in accordance with the processing contents of process 1. This is because the processing contents of the individual processes 1 are different from each other.
There are various processes such as a process mainly for requesting a memory block as small as about a byte, and a process mainly for requesting a memory block of several megabytes to several tens of kilobytes for performing a piece or image processing.

Process 1 is based on St corresponding to its own processing contents.
In order to determine the value, a request is made to the dynamic memory management unit 10 to collect a history of the allocation size at the time of the memory block reservation request, and the dynamic memory management unit 10 is called. When the process request analysis unit 21 of the dynamic memory management unit 10 recognizes that the request from the process 1 is a request for collecting a history regarding the size at the time of securing the memory block, the St storage unit 1
The flag of the history collection instruction in 1 is set, and the process returns to process 1. At this time, the history collection area in the St storage unit 10 has a reserved size range of 1 to 16 bytes, 17 to 32 bytes, 33 to 64 bytes, 65 to 128 bytes, ..., 4095 to 8192 bytes. , ...
And the like, which is composed of reserved size range identification information in the form of 2 to the power of n and a counter field for counting the number of uses for each range. When a memory allocation request is issued from the process 1, the memory allocation size analysis unit
The history collection instruction flag in the t storage unit 11 is referred to, and when the flag is set, the counter value in the range corresponding to the required securement size is increased by one.

Next, after the history collection regarding the memory reservation size is performed, when the process 1 requests the collection of the history regarding the memory block reservation size, the process request analysis unit 21 stores the history collection area in the St storage unit 11. Pass the contents to process 1. The process 1 displays the contents of the history collection area on a display device or the like, so that it can be used as a reference for obtaining an St value relating to the optimal memory allocation size for the process 1.

Further, the process 1 notifies the dynamic memory management unit 10 of an St value corresponding to the memory block size used by the process 1, and the dynamic memory management unit 10 determines a memory block allocation direction according to the St value.

The process 1 calls the dynamic memory management unit 10 by designating the St value. Dynamic memory management unit 1
0 process request analysis unit 21 analyzes the request contents of process 1 and recognizes that it is a St value setting request,
The value is set in the St storage unit 11. In the subsequent memory securing processing of the dynamic memory management unit 10, the operation is performed using the St value set in the St storage unit 11. If the St value setting request from Process 1 is not made,
Since a predetermined value (for example, 1024 bytes) is set in advance in the St value setting position in the St storage unit 11, the cutout direction can be determined based on the value.

Next, the dynamic data section 3 of the process 1
3 is a memory block allocation target area in the present invention. The start address and the end address are obtained by the dynamic memory management unit 10 issuing a system call for obtaining the address prepared by the operating system (OS). I do. At this time, the end address passed by the OS is (the amount of physically mounted memory + the size of the swap area on the disk). Therefore, if memory reservation is performed inadvertently and frequently due to the failure of the process 1 or the like, the memory of the system is occupied by the failure process because of the inconvenient operation of the failure process. May be affected. Since the process 1 can predict the upper limit value of the amount of dynamically allocated memory to be used to some extent from the history information in the above memory allocation,
The process 1 notifies the dynamic memory management unit 10 of the end address.

When the process request analysis unit 21 of the dynamic memory management unit 10 recognizes this request, the process request analysis unit 21 replaces the start address and the end address of the end address in the St storage unit 11 with the values specified by the process 1 and thereafter. In the memory allocation request, the end address is used for memory allocation from the rear side.

Next, the dynamic data section of the process 1 is divided into a plurality of sections, a start address and an end address are set for each of the plurality of divided areas, and an St value is set for each of the divided areas. Of the divided areas, the area 1 corresponds to a memory request size of 1 to 1024 bytes, and the area 2 corresponds to a memory size of 1024 to 8192 bytes, so that a more effective memory can be secured.

[0066]

As described above, according to the bidirectional memory allocation management method of the present invention, a dynamic data section existing in the space of a process is set as a target area when a memory allocation request or a release request is issued from the process. A comparison is made with the value St determined for the reservation request size Sx, and if Sx is greater than or equal to St, the requested memory block is allocated in the direction from the higher address to the lower address of the dynamic data section, and In the case where the reservation request size Sx is smaller than St, the requested memory block is allocated in the direction from the lower address to the higher address of the dynamic data section to reduce the fragmentation of the memory. Area is reduced,
As a result, there is an effect that the enlargement of the memory size of the process is suppressed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a flowchart illustrating an example of an operation according to the first exemplary embodiment of the present invention.

FIG. 3 is a block diagram showing a second embodiment of the present invention.

FIG. 4 is a schematic view of a process to which the present invention is applied.

FIG. 5 is a schematic diagram of a free memory block of the present invention.

FIG. 6 is a detailed diagram of memory reservation from a free memory block queue according to the present invention.

FIG. 7 is a diagram illustrating a memory block allocation direction when a memory is allocated according to the present invention.

FIG. 8 is a diagram illustrating a memory block securing direction when a conventional memory is secured.

[Explanation of symbols]

 1 Process 2 Memory allocation / release request analysis unit 3 Memory allocation size analysis unit 4 Free memory block search unit 51 Large free memory block management unit 52 Small free memory block management unit 61 Large memory block address space area reservation unit 62 Small memory block address Space area reservation unit 7 Requested memory block return unit 8 Memory release size analysis unit 9 Free memory block registration unit 10 Dynamic memory management unit

Claims (6)

[Claims] 1. A request for securing a memory smaller than a predetermined size (hereinafter, abbreviated as St) in response to a memory securing request from a process in an information processing apparatus. Search and secure a free area in the direction from the address to the high address,
A bidirectional memory allocation management method characterized by searching for and securing an empty area from a high address to a low address in the continuous memory space for an allocation request having a size equal to or larger than St. 2. The bidirectional memory reservation according to claim 1, wherein history information relating to the frequency of use for each request size is collected for the requested size of memory reservation from the process, and the collected information is provided. Management method. 3. The system according to claim 1, wherein a value of St is notified from a process, and a memory is allocated based on the value. 4. The bidirectional memory reservation management system according to claim 1, wherein when the value of St is not notified from a process, the memory is reserved by a default value. 5. A process for designating the size of an initial value of a continuous memory space consisting of the memory securing free area from a process, and securing the memory space having a size corresponding to the designated size. 2. The bidirectional memory allocation management system according to claim 1, wherein: 6. A continuous memory space comprising a plurality of continuous memory spaces each comprising said memory securing free area, and selecting one of said plurality of continuous spaces according to a memory securing request size from a process. For a request for securing a size smaller than St based on each defined St, a free area is searched from a low address to a high address in a continuous memory space composed of free areas for memory reservation and secured, and is equal to St. 2. The bidirectional memory allocation management system according to claim 1, wherein for a request for allocation of a size larger than that, a free area is searched and allocated from a high address to a low address in the continuous memory space.