JP2005157714A - Memory management system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a memory management system which can suppress occurrence of fragmentation of a memory area and can perform rapid allocation processing. <P>SOLUTION: The memory management system, which selects free space of the memory area for data to which allocation has been requested, and allocates the memory in an area of variable length, is provided with a reference size storage means which stores a reference size used for deciding whether the search direction in the space area is to be in an ascending order or a descending order of the memory addresses. The search direction is set in the descending order (S3) if the size of the data is equal to or more than the reference size, the search direction is set in the ascending order (S4) if the size is smaller than the reference size, and free space searched out at first as space with a size not smaller than the size of the data is selected to allocate (S5, S6) the memory for the data. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a memory management system applicable to an apparatus using a memory such as an information processing apparatus or an image processing apparatus.

  Conventionally, when an allocation request is made for a memory area, a representative algorithm for selecting one of the scattered free areas and allocating a variable-length area memory (dynamically allocating memory) In general, two types of methods are generally known: a first-fit method and a best-fit method.

  The first adaptation method searches for free areas where data requested to be allocated can be stored, in ascending order of memory addresses regardless of the size of the free area, and selects and allocates the first searched free area. It is a method to do.

  In addition, the best adaptation method searches all free areas scattered in the memory, selects free areas with a size that minimizes the difference from the size of the data requested for allocation, and allocates them. How to do it.

As another conventional technique, statistics on the size of data requested for past allocation are taken and classified according to the size, so that the allocation request size is adapted in the order of classes with the highest number of allocation requests. Some select an empty area (see, for example, Patent Document 1).
Japanese Patent Laid-Open No. 09-016463

<< Description of Background Art Operation: Specific Example 1 >>
The operation of the conventional memory management system to which the algorithm of the first adaptation method is applied will be described using a specific example shown in FIG. FIG. 7 is a diagram showing a first example of the usage state of the memory 105 in the conventional memory management system to which the algorithm of the first adaptation method is applied.

  FIG. 7A shows the memory 105 after the allocation of the data of 1 kilobyte and 32 kilobytes is alternately performed four times for each of the memory 105 from a state in which the entire area is unused (free). It shows the usage status. FIG. 7B shows the usage state of the memory 105 after deleting (releasing) the four areas to which data having a size of 1 kilobyte is allocated from the state of FIG. 7A.

  A plurality of pieces of information (data) can be stored in the memory 105, and a number for identifying a place where each piece of information is stored is given to the memory 105. This number is referred to as an address (address) or a memory address. For example, serial numbers from 0x0000 to 0x8000 in hexadecimal notation are assigned as addresses. In this case, the addresses “0x0000” and “0x8000” are referred to as the highest address and the lowest address, respectively. When two addresses are compared, the smaller number (number) is called the upper address, and the larger number is called the lower address.

  In FIG. 7A and FIG. 7B, the upper side represents the upper address side of the memory 105 and the lower side represents the lower address side. The upper end of the memory 105 is the most significant address, and the lower end is the respective upper address. Corresponds to the lowest address.

  As shown in FIG. 7A, in the first adaptation method, a free area in which data requested to be allocated can be stored is searched in ascending order of memory addresses regardless of the size of the free area. Therefore, the allocation of data having a size of 1 kilobyte and 32 kilobytes is performed alternately and continuously from the highest address of the memory 105. The area to which data is allocated is indicated as “in use” in the figure, and the area to which data of 1 kilobyte and 32 kilobytes is not allocated is an empty area (“free” in the figure). Display).

  When the in-use areas (total of 4) to which the data of 1 kilobyte is allocated are deleted, as shown in FIG. 7B, the in-use area to which the data of 32 kilobytes is allocated Four free areas having a size of 1 kilobyte are generated adjacent to the above area, resulting in fragmentation of the memory area.

  The conventional memory management system to which the algorithm of the first adaptation method is applied has been described above with reference to FIGS. 7A and 7B. However, instead of the algorithm of the first adaptation method, Even when the algorithm is applied, the memory area is fragmented as in the conventional memory management system to which the algorithm of the first adaptation method is applied.

<< Description of Background Art Operation: Specific Example 2 >>
Furthermore, the operation of the conventional memory management system to which the algorithm of the first adaptation method is applied will be described using another specific example shown in FIG. FIG. 8 is a diagram illustrating a second example of the usage state of the memory 105 in the conventional memory management system to which the algorithm of the first adaptation method is applied.

  FIG. 8 (a) initially allocates four data of size 4 kilobytes, 20 kilobytes, 8 kilobytes, 32 kilobytes, and other data to the memory 105, and then those four data, ie, the size of 4 kilobytes, 20 This shows the usage status of the memory 105 when the allocation of data of kilobytes, 8 kilobytes, and 32 kilobytes is deleted (released). It is assumed that “other data” is allocated before and after the allocation of the “four data” to the memory 105, and the allocation of these “other data” is not deleted. The areas to which these “other data” are assigned correspond to the portions described as “in use” in FIG.

  FIG. 8B shows the usage state of the memory 105 after further allocation of data of 6 kilobytes and 32 kilobytes in size from the state of FIG. In FIG. 8A, the solid line arrows and the broken line arrows indicate the order of searching for a free area in which data of 6 kilobytes and 32 kilobytes, which are requested to be allocated, are stored.

  8A and 8B, the upper side represents the upper address side of the memory 105 and the lower side represents the lower address side. The upper end of the memory 105 is the uppermost address and the lower end is the respective upper address. Corresponds to the lowest address. In addition, in the memory 105, the “4 data having a size of 4 kilobytes, 20 kilobytes, 8 kilobytes, and 32 kilobytes” that are initially allocated are allocated in the order of data having a size of 4 kilobytes, 20 kilobytes, 8 kilobytes, and 32 kilobytes. The empty areas after the respective allocations are released are hereinafter referred to as empty A, empty B, empty C, and empty D, respectively.

  In the first adaptation method, the free space where the data requested to be allocated can be stored is searched in ascending order of memory address regardless of the size of the free space, and the first free space searched is selected and allocated. Therefore, the “4 data of size 4 KB, 20 KB, 8 KB, 32 KB” that were initially allocated are allocated in this order from the upper address side of the memory 105, and the allocation related to these 4 data is After being released, the memory is used as shown in FIG. That is, empty areas are generated in the order of empty A, empty B, empty C, and empty D from the upper address side of the memory 105.

When data having a size of 6 kilobytes is further allocated from the state of FIG. 8A, a free area is searched from the higher address side as follows (the solid line arrow in FIG. 8A is indicated). reference).
(Step 1): The size of free space A (4 kilobytes) is compared with 6 kilobytes, but cannot be allocated, so the next free space is searched.
(Step 2): Since a memory area having a size of 6 kilobytes can be secured to the size of the free space B (20 kilobytes), the free space B is selected and allocated.

  Thus, for the allocation of data with a size of 6 kilobytes, regardless of whether it is optimal to select free space C (size is 8 kilobytes) with a smaller difference from the free space size, When the algorithm is applied, the vacant B is selected. Therefore, as shown in FIG. 8B, the size of the new free area (free B ′) remaining by allocating the data of 6 kilobytes to the free space B is 14 kilobytes obtained by subtracting 6 kilobytes from 20 kilobytes. Next time, when an allocation request is generated for data having a size of 20 kilobytes, the data cannot be stored in the free space B ′ (that is, the memory becomes insufficient).

Further, when allocating data having a size of 32 kilobytes further from the state of FIG. 8A, an empty area is searched from the upper address side as follows (the broken line arrow in FIG. 8A). reference).
(Step 1): The size of free space A (4 kilobytes) is compared with 32 kilobytes, but cannot be allocated, so the next free space is searched.
(Step 2): The size of free B (20 kilobytes) is compared with 32 kilobytes, but cannot be allocated, so the next free area is searched.
(Step 3): The size of free C (8 kilobytes) is compared with 32 kilobytes, but cannot be allocated, so the next free area is searched.
(Step 4): Since a memory area having a size of 32 kilobytes can be secured to the size of the free space D (32 kilobytes), the free space D is selected and allocated.

  As described above, when the algorithm of the first adaptation method is applied, all the empty areas (empty A) higher than the address where the empty D is located before the empty D in which the data of 32 kilobytes can be stored are selected. , Empty B, empty C) is determined whether or not it can be selected, resulting in waste and a great deal of time for the allocation process.

<< Description of Background Art Operation: Specific Example 3 >>
Furthermore, the operation of the conventional memory management system to which the algorithm of the first adaptation method is applied will be described using another specific example shown in FIG. FIG. 9 is a diagram showing a third example of the usage state of the memory 105 in the conventional memory management system to which the algorithm of the first adaptation method is applied.

  FIG. 9A shows the usage state of the memory 105 after a plurality of data are allocated to the memory 105 from a state where all areas are unused (empty). However, it is assumed that an empty area remains in the memory 105 even after the allocation.

  9A and 9B, the upper side represents the upper address side of the memory 105, and the lower side represents the lower address side. The upper end of the memory 105 is the uppermost address, and the lower end is the respective upper address. Corresponds to the lowest address.

  In the first adaptation method, the free space where the data requested to be allocated can be stored is searched in ascending order of memory address regardless of the size of the free space, and the first free space searched is selected and allocated. Therefore, as shown in FIG. 9A, the data is collectively allocated to the upper address side of the memory 105 and becomes “in use”, and the lower address side has an empty area (in the figure, “empty”). Is displayed).

  FIG. 9B shows the memory in the case where the allocation to the data S, data T, and data U having a size of 8 kilobytes, 20 kilobytes, and 32 kilobytes is newly performed in this order from the state of FIG. 9A. 105 shows the usage status. In the first adaptation method, the free space where the data requested to be allocated can be stored is searched in ascending order of memory address regardless of the size of the free space, and the first free space searched is selected and allocated. Therefore, as shown in FIG. 9B, the allocation to the data S, data T, and data U is performed in the area on the lower address side continuous to the area “in use” in FIG. Done in order.

  Next, whether or not fragmentation of the memory area occurs when any one of the data S, data T, and data U is deleted (released) from the state of FIG.

  First, even if only the allocation of the data U is deleted (released) from the memory 105 in the state shown in FIG. 9B, the allocated area of the data U is adjacent to the already existing empty area. The memory area is not fragmented.

  However, if only the allocation of the data T is deleted (released) from the memory 105 in the state shown in FIG. 9B, the area to which the data T is allocated is between the areas to which the data S and the data U are allocated. Therefore, fragmentation of the memory area occurs.

  Further, even when only the allocation of the data S is deleted (released) from the memory 105 in the state shown in FIG. 9B, the area to which the data S is allocated is not adjacent to the existing empty area. This will cause fragmentation of the memory area. Therefore, it can be said that fragmentation of the memory area is likely to occur when the first adaptation algorithm is adopted (as will be apparent when compared with the present invention as will be described later).

  As described above, when the first adaptation method is adopted, a free area is always searched from the beginning of the memory area, and allocation is performed as soon as a free area that can store the requested data is found. The usage frequency near the top of is higher than other parts. Accordingly, there is a problem in that fragmentation (fragmentation) of the memory area frequently occurs near the top of the memory area, and memory shortage tends to occur. Whether or not all the free areas on the higher address side than the free area can be selected even when the free area that can store the data requested to be allocated is only on the lower address side of the memory address. Since the determination is made, there is a problem that waste occurs and a lot of time is spent in the allocation process.

  In addition, when the best adaptation method is adopted, it is necessary to search all the empty areas scattered in the memory area, and there is a problem that the time spent for the allocation process increases.

  In addition, the conventional configuration example described in Patent Document 1 aims to reduce useless search processing of a free area, but when an allocation request for a size belonging to a class with a small number of allocation requests is made. After all, since it is finally explored, it takes a lot of time for the allocation process. Accordingly, when all the sizes to be selected are statistically viewed, there is a problem in the speed of the allocation process.

  In view of the above-described points, an object of the present invention is to provide a memory management system capable of suppressing the occurrence of fragmentation of a memory area and performing high-speed allocation processing.

  In order to achieve the above object, the present invention provides a memory management system for selecting a free area of a memory area and allocating a memory of a variable-length area for data requested to be allocated. Reference size storage means for storing a reference size used for determining the search direction is provided, and the search direction is set to ascending order or descending order of the memory address depending on whether the data size is larger or smaller than the reference size. While the heels are reversed, the free space first searched for as having a size equal to or larger than the size of the data is selected and the memory is allocated to the data.

  As a result, the occurrence of fragmentation of the memory area can be suppressed and high-speed allocation processing can be performed.

  Further, when there is no empty area having a size larger than the size of the data, the reference size may be rewritten and stored with the size of the data.

  Thereby, the reference size is automatically updated to a value (statistically optimal) appropriate for the program or the like that is the allocation request source.

  As described above, according to the memory management system of the present invention, it is possible to suppress the occurrence of fragmentation of the memory area and perform high-speed allocation processing.

  Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a configuration diagram of a memory management system 10 according to an embodiment of the present invention. An operating system 2 (hereinafter “CPU 1”) controlled by a central processing unit 1 (hereinafter “CPU 1”) is illustrated. This is an example in which the program 3 is running on OS2).

  When the program 3 or OS 2 performs a process that requires memory, the memory request is sent to the CPU 1 and sent from the CPU 1 to the memory management unit 4 as a memory request. In response to this memory request, the memory management means 4 allocates and secures a memory area of the requested size in the memory 5. In that case, the memory management means 4 searches for a free area in the memory 5 by referring to the management information related to the memory 5 stored in the management area 6 as a pre-process for performing allocation.

  The memory management system 10 according to the present embodiment selects one of the scattered free areas when the allocation of the memory 5 is requested to the memory 5, ie, when the allocation request is made to the memory 5, and the variable length This method employs a method of allocating memory in the area (dynamically allocating memory).

  The memory 5 can store a plurality of pieces of information (data), and a number for identifying a place where each piece of information is stored is given to the memory 5. This number is referred to as an address (address) or a memory address. For example, serial numbers from 0x0000 to 0x8000 in hexadecimal notation are assigned as addresses. In this case, the addresses “0x0000” and “0x8000” are referred to as the highest address and the lowest address, respectively. When two addresses are compared, the smaller number (number) is called the upper address, and the larger number is called the lower address. For convenience of explanation, the memory 5 and the management area 6 are illustrated separately in FIG. 1, but the management area 6 may be provided inside the memory 5.

  The memory area allocation operation of the memory management system 10 according to the present embodiment will be described below with reference to FIG. FIG. 2 is a flowchart for explaining the operation of the memory management system 10 in FIG.

  First, in step S1, it is assumed that an allocation request having a data size α is generated. In step S <b> 2 that moves through step S <b> 1, the size α is compared with the “reference size” stored in the management area 6. Here, the “reference size” is used to determine whether the exploration direction of the free area is ascending order or descending order of the memory address, as will be described later. It is determined appropriately according to the program 3 and OS 2 to be driven.

  When the size α is equal to or larger than the “reference size” (Y in Step S2), the search direction of the free area is determined as descending order (Step S3). That is, the search direction is determined so as to search for an empty area from the lowest address in the memory 5 toward the higher address side (for example, search from the memory address 0x8000 toward 0x0000). On the other hand, when the size α is smaller than the “reference size” (N in Step S2), the search direction of the free area is determined as ascending order (Step S4). That is, the search direction is determined so as to search for an empty area from the highest address in the memory 5 toward the lower address side (for example, search from the memory address 0x0000 toward 0x8000).

  Since the determination in step S2 is made every time a new allocation request occurs (every time step S1 is executed), the exploration direction of the free area can change each time.

  When the exploration direction of the free area is determined in step S3 or step S4, the free area is searched for in the determined exploration direction, and it is checked whether the size of the found free area is equal to or larger than the size α (step). S5). If the size of the found free area is equal to or larger than the size α (Y in step S5), the free area is selected and the data requested to be assigned is assigned (step S6).

  On the other hand, if the size of the searched free area is smaller than the size α, the process proceeds to step S7, and it is determined whether or not the free area has been searched over the entire area of the memory 5.

  If the search for the empty area is not yet performed over the entire area of the memory 5 (N in step S7), the process returns to the above-described step S5 to continue the search for the empty area. If the area is being searched (Y in step S7), it is determined that allocation is not possible because there is not enough free area, and the process proceeds to step S8.

  In step S8, the memory management means 4 performs error processing related to the inability to allocate, such as transmitting a signal “incapable of allocation” to the CPU 1, and rewrites the value of “reference size” with the value of size α. The significance of rewriting the value of “reference size” with the value of size α will be described later.

  As described above, in the memory management system 10, the exploration direction of the free area is set to the ascending order of the memory addresses in the case where the size α of the data requested to be allocated is greater than or equal to the “reference size” and less than the “reference size”. Selecting and allocating the first free space searched for having a size of size α or larger while reversing whether to do or descending (with different memory address ascending or descending order) Allocate the requested data.

<< Description of Operation: Specific Example 1 >>
Next, the operation of the memory management system 10 according to the present embodiment is compared with the conventional memory management system to which the algorithm of the first adaptation method described in the background art is applied, using the specific example shown in FIG. I will explain. FIG. 3 is a diagram showing a first example of a part of the contents stored in the management area 6 in FIG. 1 and the usage status of the memory 5, and the allocation of the memory area described with reference to FIG. 7 in the background art section The memory usage state when the allocation process similar to the process is executed in the memory management system 10 is shown.

  In FIG. 3 (a), similarly to FIG. 7 (a), the allocation of data having a size of 1 kilobyte and 32 kilobytes is alternately performed four times from the state in which the entire area is unused (free) with respect to the memory 5. It shows the usage status of the memory 5 after being performed. FIG. 3B shows the usage status of the memory 5 after releasing the four areas allocated with data of 1 kilobyte in size from the state of FIG. 3A, as in FIG. 7B. Yes. Further, the “reference size” stored in the management area 6 in FIGS. 3A and 3B is both 16 kilobytes.

  3A and 3B, the upper side represents the upper address side of the memory 5 and the lower side represents the lower address side. The upper end of the memory 5 is the most significant address and the lower end is the respective upper address. Corresponds to the lowest address.

  Since 1 kilobyte, which is the size of the data requested to be allocated, is the “reference size” (less than 16 kilobytes), the search direction of the free area is ascending as described above (see step S4 in FIG. 2). Since it is equal to or larger than the “reference size” (16 kilobytes), the search direction of the empty area is in descending order as described above (see step S3 in FIG. 2).

  Accordingly, as shown in FIG. 3A, four pieces of data having a size of 1 kilobyte are continuously allocated from the highest address of the memory 5 to the lower address side, and from the lowest address to the higher address side. Thus, four pieces of data having a size of 32 kilobytes are continuously allocated. In addition, one free area remains in the vicinity of the center of the memory space to which no data is allocated.

  Then, even if the area (total 4) to which the data with the size of 1 kilobyte is allocated is deleted, as shown in FIG. 3B, the four free areas with the size of 1 kilobyte have the size of 32 kilobytes. Since the data is collectively generated on the upper address side that is not adjacent to the allocated area, the memory area is not fragmented, and the existing free area is deleted from the state shown in FIG. One continuous free area that is merged with the area is generated.

  When the same processing is performed in the conventional memory management system to which the algorithm of the first adaptation method described with reference to FIG. 7 in the background art column is applied, the memory area of the memory 105 is stored as shown in FIG. Although fragmentation occurs, the memory area does not fragment in the memory 5 as described above. Therefore, in the memory management system 10 according to the present embodiment, fragmentation of the memory area is unlikely to occur, and occurrence of memory shortage due to fragmentation of the memory area can be suppressed.

  In particular, when data processing is performed by an information processing device, an input / output device, an image processing device, or the like, data having a specific fixed data size is frequently processed. For example, the data to be transferred and processed via an input / output buffer having a fixed data size is a fixed size such as 1 kilobyte, and in the case of image processing, the data processed for each line of the image is also a fixed size. It is.

  Then, when allocation requests for data having different sizes but different sizes are generated alternately (or even if they are not alternating), as shown in FIG. 3A and FIG. 7A. It can happen normally and frequently. Further, for example, when a series of image processing is completed, the data of all lines becomes unnecessary, and the allocation of a plurality of data having a certain fixed size is deleted (released) all at once, so that FIG. It can also occur normally and frequently as shown in FIG. 7B. Therefore, if the memory management system 10 according to the present embodiment is applied to the above-described information processing apparatus or the like, it can be said that it is particularly useful because occurrence of a memory shortage due to memory fragmentation can be suppressed.

<< Description of Operation: Specific Example 2 >>
Furthermore, the operation of the memory management system 10 according to the present embodiment will be described using another specific example shown in FIG. 4 while comparing with the conventional memory management system to which the algorithm of the first adaptation method is applied. FIG. 4 is a diagram showing a part of the contents stored in the management area 6 in FIG. 1 and a second example of the usage status of the memory 5, and the allocation of the memory area described with reference to FIG. 8 in the background art section The memory usage state when the allocation process similar to the process is executed in the memory management system 10 is shown.

  4A, in the same manner as FIG. 8A, four data having a size of 4 kilobytes, 20 kilobytes, 8 kilobytes, 32 kilobytes, and other data are initially stored in the memory management system 10 according to the present embodiment. This shows the usage status of the memory 5 when it is allocated to the provided memory 5 and then the allocation of those four data, that is, data of the size of 4 kilobytes, 20 kilobytes, 8 kilobytes, and 32 kilobytes is deleted (released). . Similarly to the memory 105 described with reference to FIG. 8A, “other data” is allocated before and after the allocation of the “four data” to the memory 5. ”Is not deleted. The areas to which these “other data” are assigned correspond to the portions described as “in use” in FIG.

  FIG. 4B shows the usage state of the memory 5 after further allocation of data of 6 kilobytes and 32 kilobytes in size from the state of FIG. 4A, as in FIG. 8B. . Further, the “reference size” stored in the management area 6 in FIGS. 4A and 4B is 16 kilobytes. In FIG. 4 (a), the solid line arrows and the broken line arrows indicate the order of searching for a free area in which data having a size of 6 kilobytes and 32 kilobytes which are requested to be allocated are stored.

  4A and 4B, the upper side represents the upper address side of the memory 5, and the lower side represents the lower address side. The upper end of the memory 5 is the most significant address, and the lower end is the respective upper address. Corresponds to the lowest address. In addition, in the memory 5, the “4 pieces of data having a size of 4 kilobytes, 20 kilobytes, 8 kilobytes, and 32 kilobytes” that are initially allocated are assigned in the order of data having a size of 4 kilobytes, 20 kilobytes, 8 kilobytes, and 32 kilobytes. The empty areas after the respective allocations are released are referred to as empty A, empty B, empty C, and empty D, respectively, as in the memory 105 (see FIG. 8).

  Of the “4 data of 4 kilobytes, 20 kilobytes, 8 kilobytes, and 32 kilobytes” initially allocated to the memory 5, the data of 4 kilobytes and 8 kilobytes has a size of “reference size” ( As described above, the exploration direction of the free area is ascending order (see step S4 in FIG. 2), and the data is 4 kilobytes in size from the upper address side to the lower address side, and the size is 8 kilobytes. Are allocated in this order.

  Of the 4 data of 4 KB, 20 KB, 8 KB, and 32 KB allocated initially, the data of 20 KB and 32 KB are larger than the “standard size” (16 KB). Therefore, as described above, the exploration direction of the free area is in descending order (see step S3 in FIG. 2), and the data having a size of 20 kilobytes and the data having a size of 32 kilobytes are arranged in this order from the lower address side to the upper address side. It is assigned with.

  Then, after the allocation according to the “4 data” allocated at the beginning is released, the memory usage state as shown in FIG. That is, empty areas are generated in the order of empty A, empty C, empty D, and empty B from the higher address side.

When data having a size of 6 kilobytes is further allocated from the state of FIG. 4A, since the size is less than the “reference size” (16 kilobytes), the search direction of the free area is ascending order as described above. (See step S4 in FIG. 2) As shown below, a free area is searched from the higher address side (see the solid line arrow in FIG. 4A).
(Step 1): The size of free space A (4 kilobytes) is compared with 6 kilobytes, but cannot be allocated, so the next free space is searched.
(Step 2): Since a memory area of 6 kilobytes can be secured in the size of free C (8 kilobytes), free C is selected and allocated.

  In this way, when further allocating data having a size of 6 kilobytes, in the memory management system to which the algorithm of the first adaptation method is applied, as described above (see FIG. 8B), the free space B having a size of 20 kilobytes is allocated. On the other hand, in the memory management system 10 according to the present embodiment, the free space C having a smaller difference between the size of the data requested to be allocated (6 kilobytes) and the free space size is selected.

  This is because the allocation of data having a size smaller than the “reference size” (16 kilobytes) is performed in descending order from the highest address, and the allocation of data having a size larger than the “reference size” is performed in ascending order from the lowest address. By exploring, a relatively small free area having a size smaller than the “reference size” is easily gathered on the upper address side, and a relatively large free area having a size larger than the “reference size” is easily gathered on the lower address side. This is due to the fact that

  As shown in FIG. 4B, the size of the new free area (free C ′) remaining by allocating 6 kilobytes of data to the free C is 2 kilobytes obtained by subtracting 6 kilobytes from 8 kilobytes. However, the free space B is maintained at a size of 20 kilobytes, unlike the free space B ′ (size is 14 kilobytes) in FIG. 8B, so an allocation request for the next data size of 20 kilobytes is generated. Even so, the data can be stored in the space B. Therefore, it can be said that the memory management system 10 according to the present embodiment is less likely to cause a memory shortage and the use efficiency of the memory is good.

In addition, when allocating data having a size of 32 kilobytes further from the state of FIG. 4A, the size is equal to or larger than the “reference size” (16 kilobytes). In descending order (see step S3 in FIG. 2), a free area is searched from the lower address side as follows (see broken arrow in the figure).
(Step 1): The size of the free space B (20 kilobytes) is compared with 32 kilobytes, but cannot be allocated, so the next free space is searched.
(Step 2): Since a memory area having a size of 32 kilobytes can be secured in the size of the free space D (32 kilobytes), the free space D is selected and allocated.

  As described above, in the conventional memory management system to which the algorithm of the first adaptation method described with reference to FIG. 8 in the background art column is applied when further allocating data having a size of 32 kilobytes, all four free areas are searched. However, in the memory management system 10, the search for the free area to be allocated has been completed only by searching for two free areas (free B and free D).

  This also means that the allocation of data having a size smaller than the “reference size” (16 kilobytes) is performed in descending order from the highest address, and the allocation of data having a size greater than the “reference size” is performed in ascending order from the lowest address. By exploring, a relatively small free area having a size smaller than the “reference size” is easily gathered on the upper address side, and a relatively large free area having a size larger than the “reference size” is easily gathered on the lower address side. This is due to the fact that

  In other words, when newly allocating data with a size greater than or equal to the “reference size”, the higher address side where a relatively large free area having a size greater than or equal to the “reference size” is gathered is preferentially searched. Searching for empty areas on the higher address side where empty areas with a size smaller than the “reference size” tend to gather is suppressed, and search for empty areas to be allocated at higher speed is completed.

  Therefore, it can be said that the memory management system 10 according to the present embodiment can suppress the occurrence of fragmentation of the memory area and perform high-speed allocation processing.

  In addition, when the algorithm of the first adaptation method is adopted, the use frequency of the upper address side (near the top) of the memory area becomes higher than that of other parts (lower address side etc.) as described above. Although fragmentation of the memory area frequently occurs in the vicinity of the head, in the memory management system 10 according to the present embodiment, the deviation in the usage frequency between the upper address side and the lower address side of the memory area is reduced. Therefore, it can be said that the occurrence of fragmentation of the memory area is suppressed.

<< Description of Operation: Specific Example 3 >>
Furthermore, the operation of the memory management system 10 according to the present embodiment will be described using another specific example shown in FIG. 5 while comparing with the conventional memory management system to which the algorithm of the first adaptation method is applied. FIG. 5 is a diagram showing a part of the contents stored in the management area 6 in FIG. 1 and a third example of the usage status of the memory 5, and the allocation of the memory area described with reference to FIG. 9 in the background art section The memory usage state when the allocation process similar to the process is executed in the memory management system 10 is shown.

  FIG. 5A shows the usage state of the memory 5 after a plurality of data are allocated from the state in which the entire area is unused (free) with respect to the memory 5. The plurality of data is exactly the same as the “plurality of data allocated to the memory 105 from an unused state” described in the background art section with reference to FIG. a)), among the plurality of allocated data, a size equal to or larger than the “reference size” (16 kilobytes) stored in the management area 6 shown in FIGS. 5A and 5B. And data having a size smaller than the “reference size”. Further, it is assumed that an empty area remains in the memory 5 even after the allocation.

  5A and 5B, the upper side represents the upper address side of the memory 5 and the lower side represents the lower address side. The upper end of the memory 5 is the most significant address and the lower end is the respective upper address. Corresponds to the lowest address.

  In the memory management system 10 according to the present embodiment, as described above, if the size of the allocated data is less than the “reference size”, the free areas are arranged in ascending order from the highest address (see step S4 in FIG. 2), “ If the size is equal to or larger than the “reference size”, the search direction is changed in descending order from the lowest address (see step S3 in FIG. 2), so that the upper address side and the lower address side of the memory 5 are changed as shown in FIG. The data is allocated to the two areas and becomes “in use”, and a space between the separated areas in use is a free area (indicated as “free” in the figure).

  FIG. 5 (b) is similar to FIG. 9 (b). From the state of FIG. 5 (a), the allocation to the data S, data T, and data U having new sizes of 8 kilobytes, 20 kilobytes, and 32 kilobytes, respectively, This shows the usage status of the memory 5 when this is done in this order. As shown in FIG. 5B, since the size of the data S (8 kilobytes) is less than the “reference size” (16 kilobytes), the data S is stored in the “in use” area on the upper address side of the memory 5. Since the data T size (20 kilobytes) and the data U size (30 kilobytes) are both equal to or larger than the “reference size” (16 kilobytes), they are on the lower address side of the memory 5. Data T and data U are allocated in the descending order of the memory addresses continuously to the upper address side of the “in use” area.

  Next, from the state of FIG. 5B, whether or not the memory area is fragmented when any one of the data S, data T, and data U is deleted (released) A comparison will be made with the conventional memory management system applied (see FIG. 9B).

  First, even if only the allocation of the data U is deleted (released) from the memory 105 in the state shown in FIG. 9B, the allocated area of the data U is adjacent to the already existing empty area. The memory area is not fragmented. The same can be said even if only the allocation of data U is deleted (released) from the memory 5 in the state shown in FIG.

  If only the allocation of data T is deleted (released) from the memory 105 in the state shown in FIG. 9B, the area to which data T is allocated is between the areas to which data S and data U are allocated. Therefore, even when the memory area is fragmented and only the allocation of the data T is deleted (released) from the memory 5 in the state shown in FIG. 5B, the area to which the data T is allocated is allocated to the data U. The memory area is fragmented because it is between the area that has been allocated and the area that is already in use on the lower address side of the memory 5.

  However, if only the allocation of the data S is deleted (released) from the memory 105 in the state shown in FIG. 9B, the area to which the data S is allocated is not adjacent to an already existing free area. Although the fragmentation of the area occurs, even if only the allocation of the data S is deleted (released) from the memory 5 in the state shown in FIG. 5B, the area to which the data S is allocated already exists. Since it is adjacent to the empty area, the area where the data S to be deleted is allocated is merged with the existing empty area, and the memory area is not fragmented.

  As described above, even when the same memory area is allocated to and deleted from the memory 105 and the memory 5, the memory area is less likely to be fragmented in the memory 5 than in the memory 105. This is because in the memory management system 10 according to the present embodiment, the ascending order and descending order of the exploration direction of the free area are switched according to the size of the allocated data, so that the free area is easily gathered in the center.

<< Process when memory is insufficient (step S8) >>
Next, how the “reference size” stored in the management area 6 is automatically updated to an optimum value for the program 3 or OS 2 that is the allocation request source will be described with reference to FIG. FIG. 6 is a diagram showing how the usage status of the memory 5 changes as the “reference size” stored in the management area 6 is changed.

  6A, FIG. 6B, FIG. 6C, FIG. 6D, FIG. 6E, and FIG. 6F, the upper side is the memory 105 or the memory 5. The upper address side and the lower side represent the lower address side, and the upper ends of the memory 105 and the memory 5 correspond to the most significant address, and the lower end corresponds to the least significant address. 6A to 6F are the same memory with different data storage contents, and the total storage capacity is 32 kilobytes.

  FIG. 6A shows data L, data M, data N, and data having a size of 2 kilobytes, 14 kilobytes, 4 kilobytes, and 10 kilobytes from the state in which all areas are unused (free) with respect to the memory 5. It is a figure showing the use condition of the memory 5 after performing the allocation of O in this order. Also, when assigning these data, the “reference size” is assumed to be 16 kilobytes. Then, since all the sizes of the data L, data M, data N, and data O are less than the “reference size”, the search directions of the empty areas are all in ascending order as described above (see step S4 in FIG. 2). As shown in FIG. 6A, the memory areas are continuously allocated in the order of data L, data M, data N, and data O from the most significant address. Further, on the lower address side of the area to which data O is allocated, the size of 6 kilobytes is obtained by subtracting the data L, data M, data N, and data O allocation from the total storage capacity (32 kilobytes) of the memory 5. Free space is newly generated.

  The “reference size” (16 kilobytes) as shown in FIG. 6A cannot be said to be an appropriate value in terms of use efficiency of the memory 5. Because the size of all allocated data is less than the “reference size”, the search direction of the allocation area is always in ascending order, and the same memory area allocation as when using the first adaptation algorithm is performed. It is because it ends up.

  FIG. 6B shows the usage status of the memory 5 when the memory 5 and the management area 6 are in the state shown in FIG. 6A, and an allocation request for data P (size is 8 kilobytes) is newly generated. It is a figure showing the value of "reference size." As described above, since there is only one free area having a size of 6 kilobytes in the memory 5, data P having a size of 8 kilobytes cannot be allocated. In response to this new allocation request, a state in which allocation is not possible due to insufficient free space, in other words, a state in which there is no free space having a size larger than the size of the data requested for new allocation, is referred to as “memory over”. That's it.

  Since this state is a state in which the determination result in step S7 (FIG. 2) is affirmative, the memory management system 10 performs the process in step S8 (FIG. 2). In other words, the memory management means 4 performs error processing related to the inability to allocate such as transmitting a “not assignable” signal to the CPU 1 and sets the value of the “reference size” to the data P requested to be newly assigned. Rewrite with the value of size (8 kilobytes). Therefore, as shown in FIG. 6B, the “reference size” is changed to 8 kilobytes.

  By the way, since the allocation request source program 3 and OS 2 normally request memory allocation for areas of various sizes such as one or two, sometimes several tens of types. The changed “reference size” (8 kilobytes) is unlikely to be the minimum value or the maximum value of the data size related to the allocation request that will occur in the future. Conversely, in the future, the program 3 and OS 2 that are allocation request sources will make an allocation request for data having a size larger than the changed “reference size” and an allocation request for data having a size less than this size. It can be said that there is a high possibility that it will be performed in a mixed manner.

  Therefore, after the “reference size” is changed, the memory management system 10 has a feature that the search direction of the free area is switched according to the size of the data requested to be allocated. Become.

  In other words, by rewriting the “reference size” with the size of the data requested to be allocated that triggered the “memory over”, the “reference size” becomes a more suitable value for the allocation request source program 3 or OS 2. It has been changed.

  Further, when the CPU 1 receives the “allocation impossible” signal, the CPU 1 informs the program 3 or OS 2 which is the allocation request source, information regarding “allocation impossible”, and the allocation request source has already been allocated to the memory 5. It is determined whether there is any data that can be deleted. Here, if there is no data that can be deleted, the data P cannot be allocated, but it is now determined that the allocation of the data O may be deleted (released).

  FIG. 6C is a diagram showing the usage status of the memory 5 and the value of “reference size” when the allocation of the data O is deleted from the state of FIG. From the state of FIG. 6B, a free area of 6 kilobytes and an area where data O is allocated merged into the area where data N is allocated. It is newly generated in the area on the lower address side. Data P having a size of 8 kilobytes can be stored in a new free area having a size of 16 kilobytes.

  Then, when the memory 5 and the management area 6 are in the state of FIG. 6C, a diagram showing the usage status of the memory 5 and the value of “reference size” when the data P is allocated is shown in FIG. ). Since the size of the data P (8 kilobytes) is equal to or larger than the “reference size” (8 kilobytes), the free area is searched in descending order from the lowest address, and as shown in FIG. It is allocated from the lower address side in the free area (size is 16 kilobytes) that existed in the state of (c). Then, an empty area having a size of 8 kilobytes remains in the area on the higher address side that is continuous with the area to which the data P is allocated.

  FIG. 6E is a diagram showing the usage state of the memory 5 and the value of “reference size” when the allocation of the data M and the data N is further deleted from the state of FIG. The program 3 or the OS 2 also issues a command for deleting the data M and data N. As a result, the free area (size is 8 kilobytes) that existed in the state of FIG. 6D and the area where the data M and the data N to be deleted are merged are allocated to the data L. A free area having a size of 26 kilobytes is newly generated between the allocated area and the area to which the data P is allocated.

  FIG. 6F shows the usage status of the memory 5 and the “reference” when the data Q (size is 8 kilobytes) and the data R (size is 4 kilobytes) are newly allocated from the state of FIG. It is a figure showing the value of "size". Since the data Q is equal to or larger than the “reference size” (8 kilobytes), the free area is searched in descending order. Therefore, the data Q is allocated to the area on the higher address side that is continuous with the area to which the data P is allocated. Since the data R is less than the “reference size” (8 kilobytes), the free area is searched in ascending order. Therefore, the data R is allocated to the area on the lower address side that is continuous with the area to which the data L is allocated. As a result of allocating data Q and data R, a free area having a size of 10 kilobytes remains between the area where data Q is allocated and the area where data R is allocated.

  The usage status of the memory 5 shown in FIG. 6F is less than the “reference size” for data having a size larger than the “reference size” from the lower address side, as described with reference to FIGS. The data of the size is allocated in a group from the upper address side. This is because, when “memory is over”, the “reference size” is rewritten with the data size requested for the new allocation.

  Then, as shown in FIG. 6F, if data having a size greater than or equal to the “reference size” is allocated to the lower address side and data having a size less than the “reference size” is allocated to the upper address side, a fragment of the memory area is allocated. As described above, it is possible to realize a high-speed allocation process while suppressing the occurrence of conversion.

  In addition, the “reference size” may be changed to a value that is not appropriate for the program 3 or the like that is the allocation request source in response to “memory over”. However, since “memory over”, an unnecessary allocation area deletion command, and a new allocation request frequently occur, the state where the “reference size” is not appropriate is solved immediately. That is, it can be said that the “reference size” is automatically updated to a value (statistically optimal) appropriate for the program 3 or the like that is the allocation request source.

<< Other >>
When the present invention is implemented in an image forming apparatus or the like, the size of data requested to be allocated a memory depending on the type of apparatus is determined in advance in the apparatus design stage. For example, the maximum value (maximum size) of the data requested to be allocated is 128 kilobytes, the minimum value (minimum size) is 1 kilobyte, and so on. Therefore, the initial value of the “reference size” is smaller than the maximum value so that the feature of the present invention that the search direction of the free area is switched according to the size of the data requested to be allocated from the beginning, Further, it may be set to a value larger than the minimum value.

  In the above-described embodiment, if the size of the data requested to be allocated is equal to or larger than the “reference size”, the search direction of the free area is in descending order (step S3 in FIG. 2), and if the size is less than the “reference size”, the empty space is available. The example in which the search direction of the region is set in ascending order (step S4 in FIG. 2) is given. Of course, the symbol “≧” in step S2 in FIG. 2 is changed to any of “>”, “≦”, and “<”. Also good. In any case, the above-described effects of the present invention are realized.

  That is, if the size of the data requested to be allocated is larger than the “reference size”, the exploration direction of the free area may be in descending order, and if it is less than the “reference size”, the exploration direction of the free area may be ascending order (“>”). If the size of the data requested to be allocated is larger than the “reference size”, the exploration direction of the free area may be ascending order, and if the size is less than the “reference size”, the exploration direction of the free area may be descending order (“≦≦ ”). Further, if the size of the data requested to be allocated is equal to or larger than the “reference size”, the search direction of the free area may be ascending order, and if the size is less than the “reference size”, the search direction of the free area may be set to the descending order (“<”). Equivalent).

  In the above description, the data requested to be allocated has been described using specific numerical values such as 8 kilobytes, the storage capacity of the memory 5 as 32 kilobytes, and the initial value of “reference size” as 16 kilobytes. Needless to say, the present invention is not limited to such numerical values.

  As described above, according to the memory management system of the present invention, high-speed allocation processing can be performed while suppressing fragmentation of the memory area.

1 is a configuration diagram of a memory management system 10 according to an embodiment of the present invention. 3 is a flowchart for explaining the operation of the memory management system 10 in FIG. 1. It is a figure which shows the 1st example of the usage condition of a part of content stored in the management area | region 6 in FIG. It is a figure which shows the 2nd example of the usage condition of a part of content memorize | stored in the management area | region 6 in FIG. It is a figure which shows the 3rd example of a part of content memorize | stored in the management area | region 6 in FIG. FIG. 7 is a diagram showing a state in which the usage status of the memory 5 is changed by changing the “reference size” stored in the management area 6 in FIG. 1. It is a figure which shows the 1st example of the usage condition of the memory in the conventional memory management system to which the algorithm of the first adaptation method is applied. It is a figure which shows the 2nd example of the usage condition of the memory in the conventional memory management system to which the algorithm of the first adaptation method is applied. It is a figure which shows the 3rd example of the usage condition of the memory in the conventional memory management system to which the algorithm of a first adaptation method is applied.

Explanation of symbols

1 CPU
2 OS
3 program 4 memory management means 5, 105 memory 6 management area 10 memory management system

Claims (2)

In a memory management system that allocates memory of a variable-length area by selecting a free area of the memory area for data requested to be allocated,
Reference size storage means for storing a reference size used for determining the search direction of the empty area,
It is first assumed that the size of the data is greater than or equal to the size of the data while reversing whether the search direction is ascending order or descending order of the memory address depending on whether the data size is larger or smaller than the reference size. A memory management system comprising: selecting a free area searched for and allocating a memory to the data. When there is no free space having a size larger than the size of the data,
The memory management system according to claim 1, wherein the reference size is rewritten and stored with the size of the data.

Images