JPH10116230A - Memory management device and computer-readable recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To actualize a garbage collection which is short in process time while prolonging the life of a rewrite-frequency limited type memory. SOLUTION: A stability setting part 12, sets the stability of a new object to 0 and the stability N of an object which is neither collected nor written while a garbage collecting process is performed by a specific number of times to (N+1). An object allocation part 14 allocates the objects of 0 and 1 in stability to a RAM 8 and allocates objects of 2 and 3 in stability to a flash memory 7. When there is no unused area, a garbage collection execution part 13 adds an area of higher stability from an area of 0 in stability and repeats the garbage collecting process. Thus, the frequency and time of the garbage collecting process are reduced and shortened to rewrites the flash memory 7 accompanying the garbage collection, thereby prolonging the life of the write-frequency limited type memory.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a memory management device and a computer for realizing a garbage collection with a high memory utilization efficiency and a short interruption time while extending the life of a memory having a limited number of rewrites. The present invention relates to a readable recording medium.

[0002]

2. Description of the Related Art In recent years, a PDA (Personal Di) has been developed.
A portable information processing terminal device that operates on a battery, such as a digital assistant, has been realized. A terminal such as the above PDA is strongly required to operate with a small memory capacity and to have low power consumption. In the above PDA, in order to save the state of the system when not in use, the RAM backed up by a battery (random RAM)
Access memory). Further, in order to reduce the power consumption, a considerable part of the RAM is replaced with a low power consumption flash memory.

On the other hand, garbage collection is known as a memory management method for increasing the efficiency of memory usage and facilitating the construction of application programs. For example, the operating system of Newton (made by Apple Inc.) adopts the garbage collection as a main memory management method. However, in the above Newton, the target of garbage collection is RAM
Area only.

[0004]

However, the conventional garbage collection has the following problems when used as a memory management method for a terminal system such as the PDA.

(1) Interruption of execution During execution of the garbage collection, execution of another system including an application is interrupted. Therefore, the execution of garbage collection for a certain amount of memory or more is performed on a PDA that assumes interactive use.
This will worsen the response to the user.

(2) Frequent rewriting of memory Since the above garbage collection normally presupposes memory management on a RAM, the memory space to be managed is frequently scanned and rewritten without exception. Therefore, if the garbage collection is directly applied to a memory having a limited number of times of rewriting such as a flash memory, the life of the memory is significantly shortened.

That is, in order to realize garbage collection in a computer system equipped with a memory having a limited number of rewrites, such as a flash memory, as in the above-mentioned PDA, it is necessary to suspend execution of another system by executing garbage collection. At the same time as reducing the time, it is necessary to take measures to extend the life of the memory with a limited number of rewrites.

A technique for reducing the processing time of garbage collection is called "real-time garbage collection", and many methods have been proposed. However, real-time garbage collection intended for a memory area including a memory with a limited number of rewrites for a system equipped with a memory with a limited number of rewrites such as a flash memory has not been known so far.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a memory management device which can realize a garbage collection with a high use efficiency of a memory and a short processing time while extending the life of a memory having a limited number of rewrites. To provide

[0010]

According to one aspect of the present invention, there is provided a system having a memory having a limited number of rewrites, a RAM, and an address converter for converting a logical address of both memories into a physical address. A garbage collection execution unit that performs a garbage collection process on a memory having a small unused area when the unused area of the memory with a limited number of rewrites or the RAM is smaller than a predetermined size;
When allocating a new object to each memory or executing the garbage collection process for an object which is information indicating a unit of data allocation to each memory, the garbage collection process does not collect the garbage collection process. And a stability setting unit for setting a stability that increases when writing has not been performed until the time when a new object is allocated to each memory or when the garbage collection process is performed. An object allocating unit that allocates an object whose stability is equal to or less than a predetermined value to the RAM, and allocates an object whose stability is higher than the predetermined value to the rewrite limit type memory.

In the above configuration, when a new object is requested to be assigned to the memory of limited number of times of rewriting or the RAM, the stability of the new object is set by the stability setting section. When the set stability is equal to or less than a predetermined value, the new object is allocated to the RAM by the object allocation unit. On the other hand, when the stability is higher than the predetermined value, the new object is allocated to the rewrite frequency limited memory. Further, when allocating the new object, if the unused area of the rewrite limited memory or the RAM to be allocated is smaller than a predetermined size, the garbage collection execution unit performs garbage collection processing. Then, the stability setting unit increases the stability of an object that has not been collected by the garbage collection process and has not been written until the garbage collection process. Then, when the set stability becomes higher than the predetermined value, the object is re-allocated from the RAM to the rewrite limit type memory by the object allocation unit.

In this way, while always allocating an object whose stability is equal to or less than a predetermined value to the RAM, and allocating an object having a stability higher than the predetermined value to the memory for limiting the number of times of rewriting, the garbage collection process is performed. An object group that is difficult to collect and hard to write is allocated to the memory of limited number of times of rewriting. As a result, rewriting and writing to the limited number of times of rewriting memory are suppressed, and the life of the limited number of times of rewriting memory can be extended.

According to a second aspect of the present invention, in the memory management device according to the first aspect, the garbage collection execution unit is configured to provide the high stability with respect to the memory in which the unused area is smaller than a predetermined size. It is characterized in that the garbage collection process is performed more frequently in the area where the object with the lower stability is allocated than in the area where the object is allocated.

According to the above configuration, the garbage collection process is performed with high frequency on the area to which the object with low stability is allocated. Therefore, rewriting associated with the garbage collection process is frequently performed on the RAM to which the object with low stability is allocated, and the life of the memory with limited number of rewriting can be further extended. Can be Furthermore, the garbage collection process is performed with high frequency on the area to which the object with the low stability is allocated, so that the garbage collection process is performed uniformly on the entire area of the corresponding memory. Also, the time of the garbage collection process is shortened, and the interruption time of other processes is shortened.

According to a third aspect of the present invention, in the memory management device according to the second aspect, the garbage collection execution unit is configured to determine that the unused area is smaller than the predetermined size with respect to the memory having a size smaller than the predetermined size. Until an unused area is obtained, the above-mentioned garbage collection process is performed by sequentially adding an area to which an object group having a higher stability is allocated to an area to which an object group having the lowest stability is allocated. It is characterized by:

According to the above-mentioned structure, the number of times of the garbage collection process increases in the area where the object group with the low stability is allocated, and the rewrite frequency limited memory in which the object with the high stability is allocated. For the garbage collection process is reduced. In this way, the life of the memory with limited number of rewrites can be further extended. further,
The processing time is reduced as compared with the case where the garbage collection processing is uniformly performed on the entire area of the memory in which the unused area is smaller than the predetermined size, and the interruption time of other processing is reduced.

According to a fourth aspect of the present invention, in the memory management device of the first aspect, the stability setting unit sets the stability of the new object to a minimum value when the new object is allocated. When the garbage collection process is executed, if the number of garbage collection processes performed on objects having the same stability in the process target region reaches a predetermined number of times after the garbage collection process reaches the stability, the garbage collection process is performed. The stability of an object which is not written during the predetermined number of garbage collection processes is increased by a predetermined value.

According to the above configuration, since the stability of the new object is set to the minimum value, the new object which is mostly turned into garbage that cannot establish a reference relationship with another object in a short period of time is subjected to the garbage collection processing. It is allocated to the RAM that is frequently used. In this way, the memory utilization efficiency is improved. Further, since the stability of an object to which writing is difficult is increased, an object which is predicted not to be written is assigned to the memory of limited number of times of rewriting. In this way, the life of the memory with limited number of rewrites can be further extended.

According to a fifth aspect of the present invention, in the memory management device according to the first aspect of the present invention, the address conversion unit is a page as a unit area of the address conversion on the memory with a limited number of rewrites. Access exception detecting means for detecting an access exception that occurs when a write is requested for a write-protected page,
When an access exception is detected by the access exception detecting means, the contents of the physical page on the physical address space of the memory for limiting the number of rewrites onto which the logical page on the logical address space requested to be written is mapped The RAM is copied to an unused physical page in a physical address space of the RAM, and the copied physical page of the RAM is associated with the logical page requested to be written.

According to the above configuration, when an access exception requiring a write to a write-inhibited page on the memory with a limited number of rewrites occurs, the logical page in the logical address space for which the write is requested is generated. Is mapped to a physical page on the physical address space of the RAM, so that writing to a write-inhibited page on the memory with limited number of rewrites can be performed thereafter. Furthermore, the writing that has occurred in the logical page on the memory with limited number of rewrites is performed on the RAM, and as a result, the life of the memory with limited number of rewrites can be further extended.

According to a sixth aspect of the present invention, in the memory management device according to any one of the first to fifth aspects, the memory for limiting the number of times of rewriting is a flash memory. I have.

According to a seventh aspect of the present invention, a memory management device for a system having a memory having a limited number of rewrites, a random access memory, and an address conversion unit for converting logical addresses of both memories into physical addresses. Computer-readable recording medium on which a program for storing a program for rewriting is limited, and when the unused area of the memory for limiting the number of rewrites or the random access memory is smaller than a predetermined size, A garbage collection execution unit that performs garbage collection processing, and when allocating a new object to each memory or for an object that is information indicating a unit of data allocation to each memory, Garbage A stability setting unit for setting a stability that is increased when the data is not collected by the collection processing and writing is not performed until the garbage collection processing; and when a new object is allocated to each of the memories, or when the garbage collection processing is performed. At the time of execution, an object whose stability set by the stability setting unit is equal to or less than a predetermined value is allocated to the random access memory, and an object whose stability is higher than the predetermined value is allocated to the rewrite limit memory. It is characterized by recording a program for functioning as an object allocating unit.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the illustrated embodiments. FIG. 1 is a block diagram of a computer system equipped with the memory management device according to the present embodiment. In this embodiment, it is assumed that a flash memory is used as an EEPROM having a limited number of rewrites.

The computer system 1 shown in FIG. 1 has a memory management device 2 which is a feature of the present embodiment, a memory unit 5 in which a data memory is mounted, and a program and data according to management information in the memory management device 2. Memory management unit (hereinafter abbreviated as MMU) 4 for converting a memory address (logical address) of the memory device into a memory address (physical address) on hardware;
A CPU (Central Processing Unit) 3 that controls the U4 and the memory unit 5 to perform the above-described address conversion, garbage collection, and the like.

The data memory mounted on the memory unit 5 is composed of a flash memory 7 which is a memory for limiting the number of rewrites and a readable / writable RAM 8. The memory management device 2 includes an address conversion table storage unit 9, a physical address space management unit 10, a logical address space management unit l.
1, stability setting unit 12, garbage collection execution unit 1
3 and an object allocation unit 14.

The address conversion table storage section 9 is constituted by a RAM, and stores an address conversion table indicating the correspondence between the logical addresses and the physical addresses. The address conversion table is referred to when the MMU 4 performs the address conversion. The physical address space management unit 10 manages a used area and an unused area in the physical address space. The logical address space management unit 11 manages a used area and an unused area in the logical address space.

The above-mentioned stability setting unit 12 provides information on the information (object) indicating the unit of data allocation requested by the application program, such as "the difficulty of being collected by garbage collection" and the "the difficulty of writing."
Set "stability" which is an index (numerical value) obtained by combining. Here, the “recovery” means, as described in detail later,
When performing garbage collection, an object becomes garbage and is lifted.

The setting of the "stability" is performed as follows. -Minimize the stability of the newly allocated object ("0" in the present embodiment). For an object group having a certain degree of stability, after the above-mentioned degree of stability, after the garbage collection is performed a predetermined number of times determined for the degree of stability,
The above-mentioned stability is increased (in this embodiment, by “l”). However, with respect to the objects written during the predetermined number of garbage collections, the increase in the stability is suppressed.

The garbage collection execution unit 13
Garbage collection is performed when the unused area of the data memory decreases. Further, the object allocating unit 14 allocates a new object to a logical address space, or reallocates an object whose stability has changed during the garbage collection process to a new stability area.

The computer system 1 having the above configuration operates as follows. Here, objects of low stability (stability 0 and stability 1) are set in the flash memory 7 of the memory unit 5, while objects of high stability (stability 2 and stability 3) are set in the RAM 8. Is set. First, when a new object allocation request is made by the CPU 3, the stability of the new object is set to the minimum stability "0" by the stability setting unit 12, and the logical address space management unit 1 is set by the object allocation unit 14.
1 is referred to and allocated to an unused area in the logical address space for the RAM 8. At this time, the above R
When there is no unused area in the logical address space for AM8 to such an extent that a new object can be allocated, the garbage collection executing unit 13 continuously sets an object group having a stability of “0” in the logical address space. The garbage collection process is performed on the existing area to obtain a sufficient unused area. If a sufficient unused area cannot be obtained, the garbage collection process is performed while sequentially adding areas with higher stability until a sufficient unused area is obtained. In this way, the processing time can be reduced with a smaller number of garbage collections than when the garbage collection processing is performed uniformly for all areas of the logical address space at all times, and low-stability objects that become garbage at an early stage are reduced. The frequency of the garbage collection process is increased to increase the object collection rate.

The stability setting unit 12 controls the stability N of the logical address space by the garbage collection execution unit 13.
In the area when the only garbage collection process has been performed for a predetermined number of times X _N, is that there is no the recovered a waste for reference relationships with other objects is satisfied, and the write is performed so far The stability of the missing object is increased to (N + 1). And the stability is "l"
Objects to be increased to "2" from the by the object allocation unit 14, is re allocated to the flash memory 7 side from the RAM8 side when the X _N-th garbage collection process. In this manner, the life of the flash memory 7 is extended by collecting only stable objects predicted not to be written in the flash memory 7 where rewriting is less frequently performed in the garbage collection process.

Hereinafter, the MMU 4, the stability setting unit 12,
The operations of the garbage collection execution unit 13 and the object allocation unit 14 will be described in more detail. FIG.
FIG. 3 is a conceptual diagram showing a configuration of the object. The objects are sequentially arranged in one continuous logical address space from a lower address to a higher address. The configuration of one object has the following features in order for the garbage collection execution unit 13 to execute the garbage collection process. (A) The object includes a header 21 representing the head, a number 22 of objects to be referenced, and an information storage area of pointers 23, 23,.
Each information storage area has a size of one word. At the head of the header 21, a 1-bit write flag 2
4 is set, and the mark flag 25 is set in the next one bit. Each of these flags has negative logic, and the initial value is “1”. In the write flag 24, “0” is written when writing is performed on the object (hereinafter, this is referred to as “set”). Then, it is cleared to “1” when the garbage collection process is performed on the object. Therefore, by monitoring the write flag 24, it is possible to know whether or not writing has been performed on the object during the predetermined number of garbage collections necessary for increasing the stability. When the object is referenced from another object, “0” is written to the mark flag 25 at the time of garbage collection (hereinafter, this is referred to as “set”). In the flash memory 7, "1" is written to each bit of all words after erasing data of each block, and "1" of each bit is changed to "0".
It is possible to rewrite to (the reverse is not possible). Therefore,
In the case of an object located in the logical address space set on the flash memory 7, the write flag 2
4 and the mark flag 25 can be realized. In the present embodiment, only the mark flag 25 is set for an object on the flash memory.

(B) In the pointer 23, the logical address of the head (header 21) of another object to be referred to is written. (C) The head logical address of the object can be known from the logical address indicating a word in the middle of the object. This is done by adding, to the header 21, information indicating “number” written in the number 22 of the referenced object or information indicating a pattern identifiable with the information of “pointer” written in the pointer 23 of the object, This can be achieved by searching for information exhibiting the above pattern from a logical address indicating a word to a lower address.

In this embodiment, as shown in FIG. 2, objects having the same degree of stability are continuously arranged on the same logical address space. Then, the logical address space management unit 11 manages the stability of the object group arranged in each logical address space and the range on the logical address of the area where the object group having the stability is arranged. By doing so, the stability setting unit 12
Can inquire the stability given to each object by inquiring the logical address space management unit 11 about the head address of the object group.

FIG. 3 is a conceptual diagram showing an example of address conversion performed by the MMU 4 when a write is requested by the CPU 3. As shown in FIG. 3, the logical address space includes a continuous space set for the flash memory 7 and a continuous space set for the RAM 8. Note that there is another continuous space for the ROM 6, but this is omitted in the present embodiment. Similarly, the physical address space includes a continuous space set on the flash memory 7 and a continuous space set on the RAM 8.

Usually, the address conversion by the MMU 4 is often used for "virtual storage" in which a large logical address space is mapped to a small physical address space by using a secondary storage device for backup. However, in the present embodiment, for the sake of explanation, it is assumed that no secondary storage device such as a disk is provided, and the logical address space of the RAM 8 is smaller than the set physical address space.

In FIG. 3, the object a is located on the flash memory 7 side and has an area smaller than a unit area (hereinafter, referred to as a page) P for address conversion. The object b is located on the flash memory 7 side and has an area larger than the page P. Object c is RA
On the M8 side, it has an area smaller than the page P. The object d is located on the RAM 8 side and has an area larger than the page P.

The MMU 4 mainly includes the flash memory 7
Is written on the page in the logical address space on the flash memory 7 side mapped to the page and for which "write inhibit" is entered in the address conversion table stored in the address conversion table storage unit 9. Access exception detecting means 15 for detecting an “access exception” to be performed
have. When the access exception detection unit 15 detects the access exception, the MMU 4 reads a page in the physical address space on the flash memory 7 onto which the write-protected page is mapped (hereinafter, referred to as a physical page). The abbreviated name) is copied to an unused page in the physical address space on the RAM 8, and the copied physical page on the RAM 8 is copied to a page (hereinafter referred to as a logical page) in the logical address space on the flash memory 7 side of the mapping source. ).
By doing so, it becomes possible to write to a logical page for which writing was prohibited thereafter. That is, the flash memory 7 in which the above-mentioned write prohibition is designated
Can also be used as a data memory.

FIG. 4 is a conceptual diagram showing how the address conversion is performed when the access exception occurs. In FIG. 4, write protection is designated for the first logical page of the object b in the logical address space set for the flash memory 7, and writing is requested by the CPU 3 for this logical page. MM
The state of address conversion of the object b performed by U4 is shown.

The address conversion process at the time of the above-mentioned access exception is performed as follows. FIG.
4 is a flowchart of an access exception address conversion processing operation performed by Step No. 4; Hereinafter, the address conversion processing at the time of the access exception will be described in detail with reference to FIG.

The access exception detecting means 15 of the MMU 4
When the access exception is detected, the access exception address conversion operation starts. Step S1
Then, an unused physical page is obtained from the physical address space on the RAM 8 based on the management information of the physical address space management unit 10. In step S2, it is determined whether an unused physical page has been obtained in step S1. As a result, if it can be obtained, the process proceeds to step S3, and if it cannot be obtained, the process proceeds to step S7.

In step S3, the unused physical page in the RAM 8 obtained in step S2 is replaced with the physical page on the flash memory 7 where the access exception has occurred (the physical page 3 on the flash memory 7 in FIG. 3).
The contents of 1) are copied. In step S4, the physical address space management unit 10 is notified that the physical page 31 on the flash memory 7 in which the access exception has occurred has become unused.

At step S5, the physical page 32 on the RAM 8 copied at step S3 is made writable in association with the original logical address of the physical page on the flash memory 7 that caused the access exception. The entry in the address conversion table is rewritten. In step S6, the write flag 24 of the header 21 of the object b to which the logical page 31 of the flash memory 7 in which the access exception has occurred belongs is set to "0". After that, the access exception address conversion processing operation ends.

Here, as described above, since the logical address space of the RAM 8 is smaller than the set physical address space, an extra area 33 (see FIG. 4) is generated in the physical address space. Therefore, step S
In 1, an unused physical page is obtained from the extra area 33, and the allocation between the logical address and the physical address is changed. When no unused physical pages are left in the extra area 33, it is determined in step S2 that an unused physical page cannot be acquired, and the process proceeds to step S7.

In step S7, for example, using the "LRU algorithm (Least Recent Used Algorithm)" known in the fields of "cache of CPU" and "virtual memory", it was previously obtained and used in step S1. Among the physical pages on the RAM 8 that are located (ie, allocated to the logical address space for the flash memory 7), the physical page that has not been accessed for the longest time from the present time is selected. In step S8, an unused physical page is selected from the physical address space on the flash memory 7 based on the management information of the physical address space management unit 10. In step S9, the block data of the unused physical page on the flash memory 7 selected in step S8 is erased. After that, the content of the longest unaccessed physical page on the RAM 8 selected in step S7 is copied to the erased physical page. In this way, by reusing the unused physical page on the flash memory 7, the unused memory page in the flash memory 7 which has caused the access exception registered in the physical address space management unit 10 in step S4. The physical page 31 will also be reused.

At step S10, the physical page on the flash memory 7 copied at step S9 is associated with the original logical address of the physical page on the RAM 8 selected at step S7. The entry in the conversion table is rewritten. In step S11, the physical address space management unit 10 is notified that the physical page on the RAM 8 selected in step S7 has become unused. After that, the process returns to the step S1, the physical page in the RAM 8 registered as unused in the step S1 is obtained, and the content of the physical page in the flash memory 7 in which the access exception has occurred in the step S3 is copied. When the entry in the address conversion table is rewritten in step S5 and the write flag 24 is set in step S6, the access exception address conversion processing operation is terminated.

As described above, after the access exception access conversion processing operation is completed, as shown in FIG. 4, the physical page of the logical page in which the access exception has occurred in the object b in the logical address space of the flash memory 7 side. Pages are allocated on the RAM 8. Therefore,
The writing process can be continued for the first logical page of the object b for which writing was prohibited.

Here, strictly speaking, the block size of the flash memory 7 may be different from the physical page size. In other words, the two typical sizes are the same in the range of about 1 KB to 4 KB, but the block size ≧ the page size is satisfied for the general flash memory and the MMU for the 32-bit CPU. If the block size is larger than the page size, the physical pages corresponding to the block size / page size are originally acquired in the RAM 8, but in the present embodiment, the block size is equal to the page size for simplicity. Let's say the size.

Next, the garbage collection execution unit 1
3 will be described. FIG. 6 shows a state of the logical address space when a new object is allocated on the logical address space. Here, in the present embodiment, the RAM 8
The stability of the object set in the logical address space for the flash memory 7 is set lower than the stability of the object set in the logical address space for the flash memory 7. Hereinafter, the stability of the objects arranged in the logical address space for the RAM 8 is “stability 0” and “stability 1”, and the stability of the objects arranged in the logical address space for the flash memory 7 is “stability 1”. It is assumed that "stability 2" and "stability 3".

In FIG. 6, when the CPU 3 requests the memory management device 2 to allocate a logical address to a new object, the object allocating unit 14 determines that the stability of the logical address space for the RAM 8 is zero. A memory area for a new object from the start address of the unused area 35 in the area (following the end of the used area 36) is allocated. FIG. 7 is a flowchart of the memory area allocation processing operation for the area of stability N in the logical address space performed by the object allocation section 14. In step S21,
A memory area of a new object can be allocated to an unused area of the stability N on the logical address space, and an area having a predetermined size or more is allocated to an unused area after the memory area of the new object is allocated. It is determined whether or not it remains. As a result, it is allocatable and
If an unused area equal to or larger than the predetermined size remains, the process proceeds to step S23. On the other hand, if not (ie, if allocation is not possible or no area larger than the predetermined size remains after allocation), the process proceeds to step S22. In step S22, the garbage collection execution unit 13 is started, and garbage collection processing is executed as described later in detail. After that, the process returns to the step S21, and if it is determined that the area can be allocated and the area larger than the predetermined size remains after the allocation, the process proceeds to the step S23. In step S23, a memory area for a new object is allocated from the top of the unused area of the stability N in the logical address space. After that, the memory area allocation processing operation ends.

The garbage collection processing executed in step S22 in the flowchart of the memory area allocation processing operation is not particularly limited.
Garbage collection by compaction method (Wegb
reit, B: "A generalized compactifiying garbagecol
lector ", Computer Journal, Vol. 15, No. 3, pp. 20
4-208).

FIG. 8 shows the transition of the logical address space by the garbage collection process by the above-mentioned sliding compaction method. This garbage collection process is performed according to the following procedure. (1) Marking As shown in FIG. 8A, a pointer 23 of a reference object from a specific object (hereinafter referred to as a root) 37 which is a reference of a reference relationship and is not collected as garbage. The mark flag 25 (see FIG. 2) of the referenced object is set to “0” while scanning (see FIG. 2) the reference object sequentially. The area where the route 37 exists may be an area of any stability in any logical address space. In the case of FIG. 8A, the object A is referred to by the route 37, and the object C and the object E are referred to by the object A. Therefore, the mark flags 25 of the objects A, C, and E are set, and the mark flags 25 of the objects B and D are not set.

(2) Calculation of Movement Amount As shown in FIG. 8B, objects B and D for which the mark flag 25 is not set are not used and become dust, so that the memory area of the dust objects B and D is used. Is packed with objects C and E that do not become garbage.
The movement amount of the objects C and E (that is, the change amount of the logical address) is calculated. In the case of FIG. 8B, the movement amount (the change amount x of the logical address) for packing the object C to the position of the garbage object B and the object E at the position immediately after the moved object C are set. The amount of movement (the amount of change y of the logical address) for packing is calculated.

(3) Pointer Correction As shown in FIG. 8 (c), the contents of the pointer 23 of the object A referred to in the reference relationship in the reference relationship are calculated based on the calculated values of the objects C and E of the referred side. Correct to the head logical address after moving. In the case of FIG. 8C, the content of the pointer 23 to the object E in the object A is corrected to the logical address (logical address β-variation y) at the head of the moved object E. Similarly, the content of the pointer 23 to the object C in the object A is corrected to the first logical address (logical address α-variation x) of the moved object C.

(4) Packing As shown in FIG. 8D, objects C and E that do not become dust are actually packed based on the above calculated values. FIG.
In the case of (d), the object C is moved to the position of the object B that becomes dust, and the object E is moved to a position immediately after the moved object C.

In the garbage collection process according to the sliding compaction method performed in the above-described procedure, two or three scans are required for an area of the same stability to be processed in the logical address space. Since the used areas are packed after the garbage collection process, small unused areas can be combined into one.

Here, the area to be subjected to the garbage collection process in the logical address space is the stability N
If the object referred to by the object existing in the area of the stability M (M> N) is present in the area of the stability N, the above-mentioned reference existing in the area of the stability N Must be treated as the above root during garbage collection. In order to recognize such a route without scanning the object group existing in the area of the stability M, the present embodiment is configured as follows.

That is, as shown in FIG. 9, the logical address space management unit 11 has an inter-stability reference table 16 as a management table. Then, in the marking of the garbage collection process, the pointer 23 of the reference object 38 of high stability M (= 2) on the flash memory 7 side is referred to as the referenced object of low stability N (= 0) on the RAM 8 side. When writing the first logical address of 39, the inter-stability reference table 16 of the logical address space management unit 11
Then, the head logical address of the reference object 38 and the word number of the pointer 23 in which the logical address is written are written.

By doing so, when executing the garbage collection process on the area of the stability 0 in the logical address space on the RAM 8 side, the garbage collection execution unit 13 firstly executes the garbage collection process of the logical address space management unit 11. The inter-stability reference table 16 is searched to see the pointer 23 of the specified word number of the reference object 38 at the logical address specified in the inter-stability reference table 16.
In this way, the stability M (= higher than the stability 0 of the processing target)
An object that is referenced by the reference object 38 existing in the area of (1, 2, 3) and has a stability of 0
Is searched for the referenced object 39 existing in the area of.
Then, the obtained referenced object 39 is treated as the above-mentioned route when the garbage collection is executed.

It is to be noted that the referenced object 3 referred to by the reference object 38 existing in the region having the stability M higher than the stability N of the region to be processed as described above.
The method of treating No. 9 as the above route is conventionally known (Ungar D: Generationscavenging: A non-di).
sruptive high-performance storage reclamationalgor
ithm, ACM SIGPLAN Notics vol.19, no.5, 1987).

The garbage collection execution unit 13
The garbage collection process is performed as described above. For example, as a result of performing the garbage collection process on the area of stability 0, when it is determined that the collection is not performed much and a sufficient unused area cannot be recovered. In
The garbage collection process is performed on the area of stability 0 and the area of stability 1 by adding the area of stability “1” one level higher. If a sufficient unused area cannot be recovered, a garbage collection process is performed by adding an area with higher stability. FIG. 10 is a flowchart of the garbage collection processing operation performed by the garbage collection execution unit 13 and sequentially adding higher stability areas.

In step S31, a garbage collection subroutine is executed for an area having a stability of 0 in the logical address space for the RAM 8. Step S
At 32, it is determined whether or not a sufficient unused area has been recovered as a result of the above step S31. As a result, if a sufficient unused area cannot be recovered, the process proceeds to step S33, and if recovered, the garbage collection processing operation ends. In step S33, a garbage collection subroutine is executed for the stability 0 and stability 1 areas in the logical address space for the RAM 8. In step S34, as a result of step S33,
It is determined whether a sufficient unused area has been recovered. As a result, if a sufficient unused area cannot be recovered, the process proceeds to step S35, and if recovered, the garbage collection processing operation ends. In step S35, a garbage collection subroutine is executed for the areas of stability 0 and 1 in the logical address space for the RAM 8 and the area of stability 2 in the logical address space for the flash memory 7. . Step S
At 36, it is determined whether or not a sufficient unused area has been recovered as a result of the step S35. As a result, if a sufficient unused area cannot be recovered, the process proceeds to step S37, and if recovered, the garbage collection processing operation ends. In step S37, the garbage collection subroutine is performed on the areas of stability 0 and stability 1 in the logical address space for the RAM 8 and the areas of stability 2 and stability 3 in the logical address space for the flash memory 7. Is executed. After that,
The garbage collection processing operation ends.

As described above, in the present embodiment,
The garbage collection execution unit 13 performs the garbage collection process from the lowest stability area, and if a sufficient unused area cannot be recovered, the areas having higher stability are sequentially increased until a sufficient unused area can be recovered. To continue the garbage collection process. Therefore, the number of garbage collection processes and the garbage collection processing time can be reduced as compared with the case where the garbage collection process is performed uniformly for all logical address spaces at all times, and the garbage collection process of other systems can be shortened. The execution interruption time can be shortened. As described above, the low stability areas of the stability 0 and the stability 1 are arranged in the logical address space for the RAM 8, while the high stability areas of the stability 2 and the stability 3 are the flash memory. 7 in the logical address space. Therefore, the rewriting executed in conjunction with the garbage collection is frequently performed on the RAM 8 which is not the limited number of rewriting memories, and is performed with a small number of times on the flash memory 7 which is the limited number of rewriting memories. become. Therefore, the life of the flash memory 7 can be extended. Also, paying attention to the fact that most objects tend to become garbage in a relatively short time, garbage collection processing is performed with high frequency on regions with low stability (that is, a group of objects with low stability). By doing so, it is possible to increase the collection rate of the objects and increase the use efficiency of the memory.

The stability setting section 12 calculates the number of garbage collection processes executed by the garbage collection execution section 13 for each area of stability N (N = 0, 1, 2) in all logical address spaces. Managing. Then, the number of times of the garbage collection process for the area of the stability N is a predetermined number X _N (N
= 1, 2, 3), the stability of the object existing in the area of the stability N and not subjected to the writing other than the rewriting associated with the garbage collection process and the collection is set to (N + 1). Here, in the present embodiment, as described above, objects having the same stability are arranged continuously in the same logical address space. Therefore, an object whose stability is (N + 1) in the area of stability N as described above needs to be moved to the area of stability (N + 1). Therefore, in the actual garbage collection process performed by the garbage collection execution unit 13, in addition to the procedure shown in FIG. 8, the object whose stability becomes (N + 1) after the garbage collection process is converted to the stability (N +
A procedure for moving to the area of 1) is required.

FIG. 11 is a flow chart of the garbage collection processing operation shown in FIG.
It is a flowchart of a garbage collection subroutine executed in 31, S33, S35, S37. In the flowchart of the garbage collection processing operation shown in FIG. 10, the garbage collection processing is instructed for the area having the stability of 0.
If it is determined in S2 and S34 that a sufficient unused area cannot be recovered, the garbage collection subroutine starts.

In step S41, from the route 37,
The mark flag 25 of the referenced object is set while scanning the pointer 23 and sequentially tracing the referenced object. In step S42, the movement amount of the object for filling the memory area of the object that becomes dust with the object that does not become dust is calculated. In step S43, the logical address space management unit l
With reference to the management information 1, it is determined whether or not there is an object whose stability is (N + 1). as a result,
If it exists, the process proceeds to step S44. If it does not exist, the process proceeds to step S47.

In step S44, the movement amount of one object determined to have the stability (N + 1) in step S43 is updated to the movement amount of the unused area in the area of the stability (N + 1) to the head. You. At step S45, the stability at step S43 is (N +
It is determined whether or not the update of the movement amount for all the objects determined to be 1) has been completed. as a result,
If the processing has been completed, the process proceeds to step S47, and if not, the process proceeds to step S46. In step S46, it is determined whether or not there is an unused area where the next object can be moved in the area of the stability (N + 1). As a result, if it exists, the flow returns to step S44 to shift to the movement amount update of the next object. On the other hand, if not present, the process proceeds to step S47.

In step S47, based on the movement amount calculated in step S42 or the movement amount updated in step S44, the content of the pointer 23 of the reference object is corrected to the head logical address after movement of the referenced object. Is done. In step S48, objects that do not become dust are packed based on the calculated movement amount or the movement amount updated thereafter. At this time, if there is an object to be moved in the area of stability (N + 1), the object allocating unit 14 is activated, and the relevant object is reallocated. In step S49, the above step S4
6, it is determined whether or not all the objects determined to have the stability (N + 1) in step S43 can move to the area of the stability (N + 1). As a result, if all objects can be moved, the process proceeds to step S50; otherwise, the process proceeds to step S51.

In step S50, for example, by comparing the value of (unused area after processing / unused area before processing) with a predetermined probability, as a result of step S48, it is determined whether or not a sufficient unused area has been recovered. Is determined. As a result, if a sufficient unused area cannot be recovered, step S51
If the recovery is successful, the garbage collection subroutine ends, and the process returns to the flowchart of the garbage collection processing operation shown in FIG. At step S51, the recovery flag set in the garbage collection execution unit 13 is set to "1", indicating that a sufficient unused area could not be recovered. After that, the garbage collection subroutine ends, and the process returns to the flowchart of the garbage collection processing operation shown in FIG.

As described above, in the garbage collection subroutine, if there is no unused area in which the next object can be moved in the area of stability (N + 1), updating of the movement amount is stopped. And assort them as they are. Therefore, in such a case, objects of the stability (N + 1) are mixed in the area of the stability N. However, in such a case, it is determined that a sufficient unused area cannot be recovered, and the recovery flag is set to “1”.
When returning to the flowchart of the garbage collection processing operation shown in FIG.
In S36, the recovery flag is referred to, and it is determined that a sufficient unused area cannot be recovered. as a result,
In the next step, the garbage collection subroutine is executed for the area of stability N and the area of stability (N + 1) in which objects of stability (N + 1) are mixed, and the area of stability (N + 1) is obtained. Of the unused area is recovered.

As described above, in the present embodiment,
The object has a stability setting unit 12 for setting “stability” having both “hardness of collection by garbage collection” and “hardness of writing”. A flash memory (memory for limiting the number of rewrites) 7 as a data memory and a RAM 8
The stability is “0” in the logical address space on the flash memory 7 side by the object allocating unit 14.
And the object of “1” are allocated, while the objects of stability “2” and “3” are allocated in the logical address space on the RAM 8 side.

When the CPU 3 requests writing of the objects allocated in the logical address space of the flash memory 7 and the logical address space of the RAM 8, the MMU 4 causes the address conversion table storage unit 9 to operate. The address conversion table stored in the logical address space is referred to, the address conversion is performed, and as shown in FIG. 3, the logical page of the object in the logical address space is mapped on the physical address space. Writing is performed according to the page.

When the access exception detecting means 15 of the MMU 4 detects that the write requested by the CPU 3 is an access exception, the MMU 4 reads the write-protected logical page on the flash memory 7 on which the logical page is mapped. The content of the physical page is copied to an unused page in the physical address space on the RAM 8, and the copied physical page is re-associated with the logical page on the flash memory 7 side of the mapping source. In this way, by making the write-protected logical page writable, subsequent writing to the write-protected object is performed on the physical page on the RAM 8. As a result, the number of times of writing to the flash memory 7 of the limited number of times of rewriting is reduced.

Here, when the CPU 3 requests allocation of a logical address to a new object, the stability setting section 12 sets the stability of the new object to the minimum value “0”, and the object allocation section 14 Is assigned to the head of an unused area in the area of stability 0 in the logical address space for the RAM 8. At this time, if there is no unused area for allocating a new object to the area with the stability 0, the garbage collection execution unit 13 causes the garbage collection execution unit 13 to execute the R operation with the lowest stability.
Garbage collection is performed on the area of the stability level 0 of AM8. If a sufficient unused area cannot be recovered, the garbage collection process is repeated by sequentially adding areas of higher stability until a sufficient unused area can be recovered.

Focusing on the fact that most objects have the property of becoming dust in a relatively short time, the RAM
By performing garbage collection at a high frequency on the low stability area on the 8-side, the following effects can be obtained. The number and time of the garbage collection process can be reduced, and the interruption time of execution of other systems can be reduced. Rewriting associated with garbage collection is performed a small number of times on the flash memory 7 which is a memory with a limited number of rewritings, so that the life of the flash memory 7 can be extended. The memory collection efficiency can be improved by increasing the object collection rate.

The stability setting section 12 determines the stability N
When the number of times of garbage collection processing for the area reaches a predetermined number of times X _N (N = 1, 2, 3), the write existing in the area of the stability N (write other than rewriting accompanying the garbage collection processing) is also collected. The stability of the object that has not been performed is set to (N + 1). Then, the garbage collection executing unit 13 moves the object to the area of the stability (N + 1) if there is an object whose stability is changed to (N + 1) when performing the garbage collection process on the area of the stability N. I do. Thus, by increasing the stability of a relatively stable object that is not written while the garbage collection process is performed a predetermined number of times set for each stability, the stability is as high as “2” or more. Therefore, only objects predicted to be less frequently written can be set in the flash memory 7, and the life of the flash memory 7 can be further extended.

An object that is highly likely to be written is more likely to generate dust due to a change in the referenced object. Therefore, even if the object having the stability of “2” is not collected by the garbage collection process for the predetermined number of times X2, if the writing is performed during that time, the stability does not become “3”. The garbage collection process is stopped in the area of high stability "2". In other words, even if the object is on the flash memory 7 side, the object that is likely to be written is stopped in the area where the garbage collection process is most frequent, and the generated garbage is collected by the garbage collection process. Therefore, the collection rate of the object can be further increased.

In order to function as the above-described memory management device, a program for executing the processing is recorded in a computer-readable recording medium such as a floppy disk or a CD-ROM in advance, and the computer is used as necessary. May be installed and used.

[0079]

As is apparent from the above description, according to the memory management device of the first aspect or the computer-readable recording medium of the seventh aspect, a new object for a memory with a limited number of rewrites or a RAM can be provided. When the garbage collection process is performed by the stability setting unit when the garbage collection process is performed by the stability setting unit when the garbage collection process is performed by the garbage collection process and when writing is not performed until the garbage collection process. The object allocating unit allocates an object whose stability is equal to or less than a predetermined value to a RAM, and assigns an object whose stability is higher than the predetermined value to the memory for limiting the number of times of rewriting. Because it is assigned to Hardly recovered by garbage collection process, and the write is unlikely object group is assigned to the rewritable count limited memory. Therefore, according to the present invention, the life of the memory with limited number of rewrites can be prolonged by suppressing rewriting and writing to the memory with limited number of rewrites.

Further, the garbage collection execution unit in the memory management device according to the second aspect of the present invention is arranged such that the garbage collection execution unit sets
Since the garbage collection process is performed with high frequency on the area to which the low-stability object is allocated, rewriting accompanying the garbage collection process is performed in the RAM to which the low-stability object is allocated. Will be performed frequently. Therefore, it is possible to further extend the life of the memory of the rewrite limit type. In addition, by performing the garbage collection process on a group of objects having low stability and easily becoming garbage at a high frequency, it is possible to increase the object collection rate and increase the memory use efficiency. Further, since the garbage collection process is performed with high frequency on the area where the object with low stability is allocated, the processing is performed more efficiently than when the garbage collection processing is uniformly performed on the entire area. The time can be reduced, and the interruption time of other processing can be reduced. Therefore, responsiveness to the user is improved.

Further, the garbage collection execution unit in the memory management device according to the third aspect of the present invention, the garbage collection execution unit sets an object area having the lowest stability until an unused area of a predetermined size or more is obtained. Since the garbage collection process is performed by sequentially adding the area to which the object group having the higher stability is allocated, the number of the garbage collection processes increases as the stability decreases. Therefore, it is possible to reduce the number of rewrites accompanying the garbage collection process for the limited number of rewrites memory to which the highly stable object is allocated, and to further extend the life of the limited number of rewrites memory. . In addition, the frequency of the garbage collection process for a low-stability object group that easily becomes garbage can be increased to increase the object collection rate. Furthermore, the processing time can be reduced as compared with the case where the garbage collection process is performed uniformly over the entire area of the memory where the unused area is smaller than the predetermined size, and the interruption time of other processing can be reduced.

Further, the stability setting section in the memory management device according to the fourth aspect of the present invention sets the stability of the new object to the minimum value when the new object is allocated. Can be allocated to the RAM where the number of garbage collection processes is large, and can be collected by the garbage collection process. Therefore, the use efficiency of the memory can be improved. Further, at the time of executing the garbage collection process, the stability of an object having the garbage collection process number of a predetermined number of times and not being written during the predetermined number of garbage collection processes is increased by a predetermined value. Thus, a stable object which is predicted not to be written is assigned to the memory of limited number of rewrites. Therefore, it is possible to further extend the life of the memory of the rewrite limit type.

In the memory management device according to the fifth aspect of the present invention, when the access exception is detected by the access exception detecting means, the logical page on the logical address space requested to be written is mapped. The contents of the physical page in the physical address space of the limited number of times of rewriting type memory are copied to an unused physical page in the physical address space of the RAM.
Since the physical page of the AM is associated with the logical page on the logical address space for which the write is requested, the logical page on the logical address space for which the write is requested is associated with the RA.
It can be re-associated with a physical page in the M physical address space. Therefore, thereafter, writing to the write-protected page on the memory for limiting the number of times of rewriting becomes possible. In other words, the above-described memory with limited number of rewrites can be used as a normal data memory.

Further, as a result of the above, writing that has occurred in a logical page on the memory for limiting the number of times of rewriting is performed to the RAM, and as a result, the life of the memory for limiting the number of times of rewriting is further extended. You can do it.

Further, in the memory management device according to the present invention, since a flash memory is used as the memory for limiting the number of times of rewriting, power consumption can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram of a computer system on which a memory management device of the present invention is mounted.

FIG. 2 is a conceptual diagram illustrating a configuration of an object.

FIG. 3 is a conceptual diagram showing an example of an address conversion performed by an MMU in FIG. 1;

FIG. 4 is a conceptual diagram showing a state of address conversion at the time of an access exception.

FIG. 5 is a flowchart of an access exception address conversion processing operation performed by the MMU in FIG. 1;

FIG. 6 is a conceptual diagram showing a state of a logical address space when a new object is allocated.

FIG. 7 is a flowchart of a memory area allocation processing operation performed by an object allocation unit in FIG. 1;

FIG. 8 is a diagram showing a transition of a logical address space by a garbage collection process by a sliding compaction method.

FIG. 9 is an explanatory diagram of an inter-stability reference table of the logical address space management unit in FIG. 1;

FIG. 10 is a flowchart of a garbage collection processing operation performed by a garbage collection execution unit in FIG. 1;

FIG. 11 is a flowchart of a garbage collection subroutine executed during the garbage collection processing operation of FIG. 10;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Computer system, 2 ... Memory management device, 3 ... CPU, 4 ... MMU,
5 memory unit, 7 flash memory, 8 RAM, 9 access conversion table storage unit, 10 physical address space management unit,
11: logical address space management unit, 12: stability setting unit
13: Garbage collection execution unit, 1
4 ... object allocation unit 15 ... access exception detection means 16 ... stability reference table 23 ...
Pointer, 24 ... write flag, 25 ...
Mark flag, 37 ... route.

Claims (7)

[Claims] A memory for limiting the number of rewrites and a random access memory;
A memory management device for a system having an access memory and an address conversion unit for converting a logical address of each of said memories into a physical address, wherein an unused area of said memory with limited number of rewrites or a random access memory has a predetermined size. A garbage collection execution unit that performs a garbage collection process on the memory having a small unused area when the size of the unused area is small; and a new object for each of the memories for an object that is information indicating a unit of data allocation to each of the memories. At the time of allocation of the garbage collection process, or at the time of execution of the garbage collection process, and a stability setting unit for setting a stability that increases when the garbage collection process does not collect and the writing is not performed until the garbage collection process. Each method When the allocation of new objects for Li,
Alternatively, when the garbage collection process is performed, an object whose stability set by the stability setting unit is equal to or less than a predetermined value is allocated to the random access memory, and an object whose stability is higher than the predetermined value is rewritten. A memory management device comprising an object allocating unit for allocating to a limited-time memory. 2. The memory management device according to claim 1, wherein the garbage collection executing unit is configured to execute the garbage collection with respect to a memory in which the unused area is smaller than a predetermined size than an area in which the object with high stability is allocated. A memory management device for performing the garbage collection process with high frequency on an area to which the object with low stability is allocated. 3. The memory management device according to claim 2, wherein the garbage collection execution unit obtains an unused area equal to or larger than the predetermined size for a memory having the unused area smaller than a predetermined size. Memory, wherein the garbage collection process is performed by sequentially adding an area to which an object group having a high stability is allocated to an area to which an object group having the lowest stability is allocated Management device. 4. The memory management device according to claim 1, wherein the stability setting unit sets the stability of the new object to a minimum value when allocating the new object, and sets the stability when executing the garbage collection process. When the number of garbage collection processes performed on the object having the same stability in the processing target area reaches a predetermined number of times after the stability, the object in the processing target area and the predetermined A memory management device characterized in that the stability of an object that has not been written during the number of garbage collection processes is increased by a predetermined value. 5. The memory management device according to claim 1, wherein the address conversion unit is a page as a unit area of the address conversion on the memory for limiting the number of times of rewriting and which is write-protected. Access exception detecting means for detecting an access exception that occurs when a write is requested to the logical address space where the write is requested when the access exception is detected by the access exception detecting means. Copy the contents of the physical page on the physical address space of the memory for limiting the number of rewrites onto which the logical page is mapped to an unused physical page on the physical address space of the random access memory, The physical page of the random access memory is associated with the logical page requested to be written. A memory management device, comprising: 6. The memory management device according to claim 1, wherein the memory for limiting the number of times of rewriting is a flash memory. 7. A memory having a limited number of rewrites and a random access memory.
A computer-readable recording medium storing a program for functioning as a memory management device of a system having an access memory and an address conversion unit for converting a logical address of each of the memories into a physical address, wherein the number of times of rewriting is limited. When an unused area of the memory or the random access memory is smaller than a predetermined size, a garbage collection execution unit that performs a garbage collection process on the memory having a small unused area, and a data allocation unit for each of the above memories. At the time of allocating a new object to each of the memories or executing the garbage collection process, the object which is not represented by the garbage collection process and written until the garbage collection process. And stability setting unit for setting a stability that increases when the write is not performed, when the allocation of a new object for each memory,
Alternatively, when the garbage collection process is performed, an object whose stability set by the stability setting unit is equal to or less than a predetermined value is allocated to the random access memory, and an object whose stability is higher than the predetermined value is rewritten. A computer-readable recording medium in which a program for functioning as an object allocating unit for allocating to a limited-time memory is recorded.