JP2002049525A - Main storage management system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To use the entire area widely and equally even when the capacity of a main storage increases and to suppress the capacity of a table regarding main storage management. SOLUTION: The management table is provided which shows the use states of blocks of large capacity while the main storage is divided into the blocks and each management block is provided with bits as many as all pages to which one bit each showing the use state of the corresponding page is assigned while the corresponding block is divided into the pages of small capacity. A main storage management part which specifies and allocates different blocks in order at requests for the allocation of the mains storage area retrieves a management table corresponding to a specified block at a request to obtain a usable page position and areas for a necessary number of pages, and updates the states of each bit of the pages in the corresponding management table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a main storage management system for dynamically managing allocation and return of a program execution area and a data area in a computer system.

[0002] In recent years, with the demand for large-scale and large-capacity computer systems, the capacity of main memory has been steadily increasing. In the conventional main memory management method,
This causes an uneven allocation area and an increase in the capacity of the management table. To prevent this, the main memory must be managed efficiently.

[0003]

2. Description of the Related Art FIG. 9 is an explanatory view of a conventional example.
0 is a main memory and 81 is a management table. The main memory 80 of the information processing apparatus of the conventional example shown in FIG. 9 is managed by dividing the whole in units of fixed storage capacity. In this description, one unit (called a page) is 4 KB (kilobytes).
It is assumed that the entire main memory 80 has a capacity of n pages (for example, n = 250). The management table 81 for managing the main memory 80 stores an identification code (represented by A and B in the example of FIG. 9) of a task using the page corresponding to each page of the main memory 80. If the number of tasks is 256 or less, the identification code length is composed of 1 byte (8 bits). In the area of the management table 81 corresponding to an unused (empty) page of the main memory 80, information for displaying the empty is set.

When a task requests a use of the area of the main memory 80 by the processing of the program in the main memory 80, the management table 81 is searched by the control of an operating system (OS) (not shown). When a free area following the used area is found, the number of pages (a multiple of 4 KB) that satisfies the requested capacity is acquired. Write. Less than,
In accordance with a request from each task, necessary storage areas are sequentially allocated in page units in a similar manner. In the example of FIG.
The first three pages (4 KB × 3 = 12 KB) are used by task A, and the subsequent six pages are used by task B. When the use of the storage area is completed and released, the area corresponding to each page of the management table 81 is set and returned with information indicating an unused state.

[0005]

In recent years, the capacity of the main memory has tended to increase. However, according to the conventional technique described with reference to FIG. However, when they are assigned to each task, they are returned to the unused state when their use ends, and the storage area near the head is repeatedly used, so that the entire main memory is not used evenly.

On the other hand, the normality of the storage device can be checked by writing and reading the entire storage area. According to the above-described conventional main storage management method, the capacity of the management table is reduced. However, there is a problem that the normality of the main memory cannot be checked because only a partial area of the main memory is used frequently. Further, since a predetermined area (1 byte in FIG. 9) is used to store the identification code of the program used in each page unit, there is a problem that the capacity of the management table also increases in response to the increase in the capacity of the main memory. Was.

SUMMARY OF THE INVENTION An object of the present invention is to provide a main memory management system which can use the entire area widely and evenly even if the capacity of the main memory increases, and can suppress the capacity of a table for management.

[0008]

According to the present invention, even if the main storage capacity is increased, a plurality of management tables are used to change the management table for each area allocation, thereby eliminating the bias of the allocated area. The above problem was solved by managing the data with 1-bit data and separately managing the relationship between the assigned address and the area.

FIG. 1 is a diagram showing the principle configuration of the present invention.
In the figure, 1 is a main memory of a large capacity x in GB (gigabyte) units, 1-0, 1-1,...
(N + 1) blocks divided by a capacity y of about (megabyte) unit, and 2-0 to 2-n are provided individually corresponding to the blocks 1-0 to 1-n obtained by dividing the main memory 1. , Is a management table in which each block is divided into small-capacity pages (for example, on the order of several KB), and one bit is allocated in the arrangement order of each page to indicate the used / empty state. 3
Is a main memory management unit, and 3a is a table search means. Each time an allocation request is issued, one of a plurality of management tables is allocated in order, and a free area is searched by the management table to perform allocation. 3b is a table setting means. Reference numeral 3c denotes a number holding unit that holds a number next to the block number (management table number) for which the area was previously allocated. Reference numeral 4 denotes a page position table in which usable page positions in each management table are set. Reference numerals 4-0 to 4-n denote each management table 2-0.
Page positions corresponding to .about.2-n. Reference numeral 5 denotes an address management table effective for managing the return of the storage area, which includes the position of the used storage area, the number of used pages, and the identification code of the program (task) using the area.

When the main memory management unit 3 is started by a storage area allocation request from a program, the table search unit 3a operates the management table 2 provided for each area obtained by dividing the main memory 1 into blocks. The number holding unit 3c stores the block number of the current area allocation from 0 to 2-n.
From the management table 2-0 to 2-n, the bit of one management table (referred to as 2-1) corresponding to the number taken out is searched, and after "1" (indicating busy) is retrieved. When detecting that “0” is set continuously,
By identifying the head bit position, the "bit position" obtained by searching the block of the detected number in the main memory 1 can be obtained as the page position, and allocation is requested from the corresponding position. The number of pages is assigned, and the table setting means 3b sets bits of the assigned number of pages of the management table 2-1 from "0" indicating unused to "1" indicating used. Also, the number holding unit 3
The value of c is updated (+1), and when a next storage area allocation request occurs, area allocation is performed for the next block (management table number). When the use of the area has been completed, the state of the corresponding page in the management table 2 is set from in use to free in order to release the allocated main storage area.

The page position table 4 includes management tables 2-0 to 0-n corresponding to the blocks 1-0 to 1-n.
2-n, the start position (address) of each free page is set, and the management tables 2-0 to 2-2 are set.
It is only necessary to detect the page position corresponding to the management table number from the page position table 4 without the trouble of searching for the content of −n. When the storage area is allocated, the page position corresponding to the corresponding management table in the page position table 4 is updated (added) by the number of allocated pages.
Further, when the storage area of the main memory 1 is allocated, the identification code of the allocated program (task), the address of the allocated page and the number of pages are stored in the address management table 5. A large number of address management tables 5 are provided for each program (task) to be assigned.

[0012]

FIG. 2 shows a configuration example of a management table corresponding to a specific example of a main memory. In this example, the main memory 1 has a capacity of 2 GB (gigabytes) (x = 2 × 2 in FIG. 1).
³⁰ B (byte)), one block is 16 MB
(Megabyte) capacity (y = 16 × 2 ²⁰ B in FIG. 1)
Case). In this case, the main memory 1
Since 2 GB / 16 MB = 128, it can be divided into 128 blocks (n = 12 in FIG. 1).
7) Therefore, a total of 128 management tables represented by reference numerals 2-0 to 2-127 are provided.

Each block is divided into 4 KB (more precisely, 409
6B), 16M
B / 4KB = 4096 pages, one block is 40
When the management table is composed of 96 pages and the status of each page (in use / empty) is represented by one bit in the management table, each of the management tables 2-0 to 2-127 corresponding to each block has a capacity of 4096 bits (512 B). Becomes

FIG. 3 shows an example of the structure of a page position table indicating the position of an area in the management table. FIG. 4 shows an example of the structure of an address management table showing the relationship between the assigned addresses and the tasks to which the tasks have been assigned.

In the case of the configuration example of the management table shown in FIG. 2, each management table has 4096 bits, and further has 12 bits.
The address length for indicating which bit position (page number including the management table number) in the entire eight management tables is required is 2 bytes (16 bits). The page position table of FIG.
2 bytes representing each page position of 0～2-127 is stored, is a total of 2B × 128 = 256B, each page position in FIG. 3 is displayed in the four-digit number of hexadecimal notation (represented by 0 _x) Have been.

An address storing information necessary for investigating a state in which a program (task) using the area of the main memory forgets the process of releasing the area after use and returning the area to its original state (making it a free area). FIG. 4 shows a configuration example of the management table. This address management table is generated for each storage area allocation, and the allocated address (4
Byte), identification ID (2
Bytes), the number of pages used in the allocation (2 bytes)
It is. The capacity of the address management table is m × 8 bytes, where the maximum number assigned is m.

FIG. 5 shows the relationship between the management table and the page position table. FIG. 5 shows an example of the page position 4-n in the page position table 4 corresponding to one of the management tables 2-n in the management table shown in FIG.

The management table n is 4096 as described above.
It is possible to manage a 16 MB block with bits. In a state where no area is allocated in this block, all are "0". When an area allocation request is issued, a method of searching for an area sequentially from address 0 is generally used, but the search is performed from a page position described in a page position table according to the present invention.

FIG. Indicates a state in which “8” is set as the page position 4-n corresponding to the management table n in the page position table 4. In this state, it is assumed that the number of pages requested from the user program is "4" and the management table 2-n has been allocated. Thereby, B. of FIG. As shown in (2), by securing an area for the fourth page from the eighth in the management table 2-n, the corresponding position in the management table 2-n is rewritten from "0" to "1", which corresponds to this. The value of the page position 4-n in the page position table is also 8 + 4 = 12, and when there is a next acquisition request, the search is performed from the twelfth position.

As described above, the page position table indicates the position at which the search of the management table is started. When the number of requested pages has been acquired, the number of requested pages is updated. Therefore,
Since it is not necessary to search from the top of the management table in the next search, the search time is reduced.

FIG. 6 is a flow chart of the process for allocating and releasing the main storage area. In the figure, reference numerals 30 to 33 denote software corresponding to the main memory management unit 3 in FIG. 1, 30 is a user program, 31 is an SVC (Supervisery Call) interrupt handler, 32 is a memory acquisition process, and 33 is a memory acquisition process. This is memory release processing.

When a request for allocating a main storage area occurs in the user program 30, the user program may use the parameter area in the fixed area (prefix) 30a which is commonly referred to by the OS (operating system) and the user program. The storage area data that can be set is set (S1 in 30 in FIG. 6). This processing is performed so that storage areas do not conflict when operating with a multiprocessor. Next, an SVC for acquiring memory is issued (S2 in 30 in FIG. 6). As a result, the SVC interrupt handler 31 is driven, a process for decoding the SVC code is performed (S7 in 31 in FIG. 6), and it is determined whether the process is a memory acquisition process (S8). The process proceeds to (S9). Memory acquisition processing 32 of FIG.
Then, data is obtained from the parameter area of the fixed area 30a (S12 in 32 in FIG. 6), and the area is obtained (S1 in FIG.
3), the acquired address is set in the parameter area of the fixed area (S14), and the process returns to the SVC interrupt handler (S14).
15). The area acquisition processing is performed by the main memory management unit by the method shown in FIG. 5 using the management table and the page position table shown in FIGS. After the area is obtained, registration in the address management table of FIG. 4 is performed.

The SVC interrupt handler 31 returns to the user program when the memory acquisition processing is completed (S11 in 31 in FIG. 6). In the user program, the data in the parameter area of the fixed area is saved to the area of the user program. Then (S3 in 30 in FIG. 6), the process is performed (S4), data is set in the parameter area of the fixed area (S5), and a memory release SVC is issued (S6).
As a result, the SVC interrupt handler 31 is activated,
The VC code is decoded (S7 in 31 in FIG. 6), and it is determined whether the decoding result is a memory acquisition process (S8).
It is determined to be no, and the process proceeds to the memory release process (S10). In the memory release processing, data of the acquisition address set at the time of acquisition is obtained from the parameter area of the fixed area (S16 of 33 in FIG. 6), and the area is released (S1 in FIG. 6).
7), return to the SVC interrupt handler (S1)
8). Also in the area release processing, the main memory management unit performs control to return to the unused state with respect to FIGS. SVC
Returning from the interrupt handler to the user program ends the memory release processing.

A running test of a large-capacity main memory can be efficiently performed over a wide range using the management table and the page position table according to the present invention.

FIG. 7 is a processing flow for acquiring and releasing an area in the running test, and FIG. 8 shows the relationship between the management table and the page position table in the running test.

The area given to the test program is determined by the current management table position (0 to n) and the page position corresponding to each management table. First, 1 is entered as the loop number (S1 in FIG. 7), and the management table is initialized (S2). Next, the current position of the management table is determined by the loop number (S3 in FIG. 7). As a result, the block number to be tested is determined from the plurality of blocks, and in this example, is determined in the management table 1 corresponding to the loop number 1. Subsequently, all the page position tables are initialized with the loop number (1 at this time) (S4 in FIG. 7), and a necessary area is secured (S5). The necessary area can be arbitrarily specified, and a plurality of areas (areas at a plurality of continuous places in the same management table) are secured.

Subsequently, a test such as read and write is executed for the secured area (S6 in FIG. 7), all the areas used in the test are released (S7), and the test for the target area is performed. It is determined whether the process has been completed (S8), and if not completed, 1 is added to the loop number (S9), and the process returns to S3. Thereby, for the management table 2, a necessary area is secured by a corresponding page position table (initialized to 2 in step S4), and a test is performed. Is performed. When the test is completed, all the areas are open. Therefore, the current management table position and the page position are determined from the loop number, so that if the loop number is the same, the address of the acquired area is also the same. Become.

FIG. 8 shows an initial state of the above-mentioned running test. That is, in the first cycle in which the loop number is set to 1 in S1 of FIG. 7, all the page position tables 4-0 to 4-n are set to 1 and the management tables 2-0 to 2-n are set. Also, a management table 2-1 is set according to the loop number (first set to number 1), and a test is performed by securing an area using the value 1 of the page position table 4-1 corresponding to the management table 2-1. . In addition, all areas are initialized for each cycle, and the positions of the management table and the page position table are uniquely determined for each block according to the loop number, and the area can be secured and the test content can be changed. . If an error is detected in the test, the same condition can be reproduced by returning to the same loop number and performing the test.

(Supplementary Note 1) The main memory is divided into large-capacity blocks, and a management table indicating the use state of the blocks is provided corresponding to each block. Each management table stores the corresponding blocks in units of small-capacity pages. A main memory management unit that is provided with bits for all pages in which one bit indicating the use state is allocated to each page, and sequentially specifies different blocks in response to a main storage area allocation request. The main memory management unit searches the management table corresponding to the designated block in response to the request, and obtains an available page position and an area for the required number of pages,
A main storage management method characterized by updating the state of each bit of a corresponding management table page.

(Supplementary note 2) In Supplementary note 1, there is provided a page position table storing the assignable page positions in each management table provided corresponding to each block of the main memory, and the main memory management unit A main storage management method characterized by detecting the page position table corresponding to a block specified by an allocation request and acquiring an area for the requested number of pages.

(Supplementary note 3) In any one of Supplementary notes 1 and 2, when a storage area is allocated in response to the main memory allocation request, an identification code indicating the program or task that has made the request is assigned to the storage area. A main storage management method, wherein an address management table including a page position and the number of pages is generated for each storage area allocation.

(Supplementary Note 4) The main memory is divided by a predetermined large-capacity block, and a management table indicating the use state of the block is provided corresponding to each block. A page position table storing the positions of allocatable pages in each management table, wherein a bit is set for all pages in which one bit of use status is allocated to each page by dividing each page into To perform a running test, determine the loop number,
A main storage management method characterized by determining a position of a management table and a page position table in the management table from a loop number uniquely determined, and uniquely determining an assigned address.

[0033]

According to the present invention, when dynamically allocating and returning a program execution area and a data area in a computer system, it is possible to eliminate the bias of the allocated area and efficiently use the main memory. it can.

Further, the capacity of a management table used for dynamically managing the main storage can be reduced. More specifically, in the case of the conventional system in which a 2 GB (gigabyte) space is divided by 4 KB and one byte of management data is provided for each 4 KB, 2 GB / 4 KB = 512
According to the present invention, 64 KB (= 2 GB / 4 KB / 8 B) is used in the management table, 256 B (2 GB / 16 MB × 2 B) is used in the page position table, and 64 KB (8192 × 8 bytes) is used in the address management table. , 129 KB in total, making it possible to reduce about ４ of the conventional method.

Further, when the running test of the main memory is performed according to the present invention, it is possible to uniquely determine the management table and the page position in the management table.

[Brief description of the drawings]

FIG. 1 is a diagram showing the principle configuration of the present invention.

FIG. 2 is a diagram illustrating a configuration example of a management table corresponding to a specific example of a main storage;

FIG. 3 is a diagram illustrating a configuration example of a page position table indicating a position of an area in a management table.

FIG. 4 is a diagram illustrating a configuration example of an address management table indicating a relationship between an assigned address and an assigned task;

FIG. 5 is a diagram illustrating a relationship between a management table and a page position table.

FIG. 6 is a diagram showing a processing flow of main memory area allocation and release.

FIG. 7 is a diagram showing a processing flow of acquiring and releasing an area in a running test.

FIG. 8 is a diagram showing a relationship between a management table and a page position table in a running test.

FIG. 9 is an explanatory diagram of a conventional example.

[Explanation of symbols]

 1 Main memory 1-0 to 1-n Block 2-0 to 2-n Management table 3 Main memory management unit 3a Table search unit 3b Table setting unit 3c Number holding unit 4 Page position table 4-0 to 4-n Page position 5 Address management table

Claims (3)

[Claims] 1. A main memory is divided into large-capacity blocks, and a management table indicating a use state of a block is provided corresponding to each block. Each management table divides the corresponding block into small-capacity page units. And a main memory management unit for sequentially allocating by designating different blocks in response to a request for allocating a main storage area. The main memory management unit searches the management table corresponding to the designated block in response to the request, acquires an available page position and an area for the required number of pages, and obtains a page of the corresponding management table. A main memory management method characterized by updating the state of each bit of a page. 2. A page position table according to claim 1, wherein an assignable page position in each management table provided corresponding to each block of said main storage is stored.
The main storage management method according to claim 1, wherein the main storage management unit detects the page position table corresponding to the block specified by the allocation request, and acquires an area corresponding to the requested number of pages. 3. A main memory is divided by a predetermined large-capacity block, and a management table indicating a use state of the block is provided corresponding to each block. Each management table stores the corresponding block in a small-capacity page. A page position table is provided which stores the positions of allocatable pages in each management table in which bits are set for all pages in which a use state of 1 bit is allocated to each page, and the main memory is run. A main memory management method characterized by determining a loop number for a test, determining a position of a management table and a page position table in the management table from a uniquely determined loop number, and uniquely determining an assigned address. .