JPH03141442A - Task control system - Google Patents

Abstract

PURPOSE:To prevent the deterioration of response to the user input by performing the compaction processing with repetition of the transfer of units that is uninterruptable and processed in a short time and performing the unit transfer processing in an idle state of a system. CONSTITUTION:A data processor allocates dynamically plural tasks into a main memory 2 and saving/recovering these tasks between the memory 2 and the secondary storages 3 and 4 to perform the multi-task processing. Then the compaction processing is carried out to collect the idle areas of the memory in a single place with repetition of the transfer of units that is uninterruptable and processed in a short time. The unit transfer processing is carried out in an idle state of a system of the data processor. Thus it is possible to prevent the deterioration of response to the user input and to perform the compaction processing with the reduced overhead.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a task control method for realizing a multitasking method in a data processing device that performs processing through interactive operations.

[Conventional technology]

Conventionally, as a method of using computer systems in data processing devices such as word processors and combinators,
A multitasking method is widely adopted in which the processing to be performed by a system is divided into independent processing units called tasks, and multiple tasks are operated in parallel, thereby processing the entire system.

In this case, instead of keeping all the contents of each task in the main memory at all times, the main memory capacity can be increased by saving and restoring the contents of each task to and from a secondary storage device such as a hard disk as needed. It becomes possible to run multiple tasks that require memory capacity that exceeds the constraints of .

As described above, the methods for saving/recovering tasks can be roughly divided into the swap method and the virtual memory method based on the processing unit for saving. In the swap method, the unit of save/restore processing is a task, and the task being executed always exists entirely in the main memory. On the other hand, in the virtual memory method, the unit of evacuation/recovery processing is a memory fragment called a page that is smaller than a task, and just because a task is being executed does not mean that the entire task is stored in main memory. It doesn't exist.

Of the above two methods, the virtual memory method has higher main memory usage efficiency, and compared to the swap method, the main memory
The amount of data transferred to and from the storage device.

Since the number can be reduced, processing can be performed faster. However, in order to implement the virtual memory method, it is necessary to provide an expensive memory management mechanism as hardware, so it is not necessarily a good method in terms of cost performance.

In contrast to the virtual memory method, the swap method does not require an expensive memory management mechanism, but it is necessary to allocate main memory that can store the entire task to one task. The size of a task is not constant, and if tasks of different sizes are repeatedly allocated/released from main memory, a state called fragmentation occurs in which free areas of different sizes are scattered in main memory, and the free space increases. Although the total amount of area is large enough, the size of each free area is smaller than the size of the task, so tasks cannot be placed in memory, reducing memory usage efficiency and The time required for evacuation/recovery processing increases and performance deteriorates.

Therefore, in the swap method, as a way to prevent performance degradation due to the above fragmentation, tasks are moved in main memory and the free area is consolidated into one continuous location. It is known to cure the condition. An example of an apparatus for carrying out such a compaction treatment is the apparatus described in Japanese Patent Application Laid-Open No. 60-97444.

[Problems to be Solved by the Invention] The compaction processing described above requires a large amount of C to move a large number of tasks by performing inter-memory transfer processing on the main memory.
Requires PU processing time. Further, while a task is being moved in main memory, the contents of the task are only partially correct and the task cannot be executed.

Since the compaction process is performed on the entire main memory, no normal task processing is performed during the compaction process, and the system concentrates on the compaction process.

Therefore, if the above compaction processing is implemented in an interactive data processing system such as a word processor, no processing requests from the user will be accepted during the compaction processing, so this is the most important performance when performing interactive processing. There was a problem in that the responsiveness to the user's human power deteriorated.

SUMMARY OF THE INVENTION An object of the present invention is to realize compaction processing with little overhead in an interactive processing system, and further, to realize compaction processing that does not deteriorate responsiveness to user input.

[Means to solve the problem]

In order to achieve the above object, the present invention dynamically allocates a plurality of tasks in main memory, performs the saving/recovery processing of the tasks between the main memory and the secondary storage device,
In a data processing device that performs multitasking processing, the above-mentioned free areas in the main memory are grouped together in one place K.
One-time processing is realized by repeating unit transfer processing, which is a short-time process that cannot be interrupted, and the unit transfer processing is performed when the system of the data processing device is in an idle state.

In other words, the unit transfer process is a process of moving one of the tasks in the main memory, and in normal mode, the activation condition for this unit transfer process is that the system remains idle for a certain period of time or more. .

Preferably, this fixed time is a value that can be determined to be a response waiting time for the user.

To this end, one aspect of the present invention is to dynamically allocate a plurality of tasks within the main memory, and between the main memory and the secondary storage device.
In a data processing device that performs multitasking processing by performing the task save/recovery process, a first table (hereinafter referred to as a task management table) that manages the state of the task and the allocation state of the task in the main memory is provided. )
a second table (hereinafter referred to as page management table) that manages the usage state of the main memory, and a fifth table (hereinafter referred to as the "page management table") that manages the save state of the task to the secondary storage device. means for moving one of the tasks in the main memory in order to consolidate free areas in the main memory into one place using the task management table (referred to as a management table), the task management table, and the page management table; and a means for performing allocation/release processing of the task in the main memory using the task management table, the page management table, and the evacuation management table; Means for saving/recovering the task to/from the secondary storage device using a management table;
In the priority mode, the releasing means controls the execution of the moving means and the means for activating the moving means when the task is not allocated in the main memory, and the task to be executed is stored in the main memory. If it does not exist, start the above allocation/release means and the above evacuation/recovery means,
Further, if a state in which there is no task to be executed continues for a certain period of time or more, means for activating the above-mentioned moving means is provided.

Specifically, the allocation/release means and the save/recovery means are realized by a task called a swap task, and the movement means is realized by a task called a GC (garbage collection) task.

Further, other means are realized as part of O8 (operating system).

[Effect]

The task management table manages the execution status of tasks and memory allocation status.

The main memory is managed by being divided into fixed-length management units called pages, and a page management table manages the usage status of each page.

The save management table manages tasks saved on the secondary storage device and their positions.

The swap task is started by the O8, checks/updates the three management tables mentioned above, allocates/releases the task to the main memory, and saves/recovers the task to the secondary storage device. , if the task cannot be allocated to the main memory, the GC task is activated to perform compaction processing in the main memory.

The GC task is started by the O8 or the swap task, refers to the task management table and the page management table, and clears the portion of main memory that is not allocated to the task.
Put the pieces together in one place (process the comb).

Once activated, the GC task moves at most one task within the main memory and finishes processing.

Compacting the entire main memory is achieved by repeatedly activating the GO task.

O8 controls the entire system, and in particular, in the task control part, if the task to be executed does not exist in the main memory, it starts a swap task and restores the task to be executed in the main memory. then run the task.

The O8 also manages the duration of an idle state in which no processing is performed by the system, and when the idle state continues for a certain period of time, it starts a QC task to perform main memory compaction processing.

In this way, the O8 detects the idle state of the system and performs compaction processing, so that the main memory compaction processing can be performed without putting a burden on the system processing capacity. Main memory is used more efficiently, increasing system processing power.

In addition, the amount of data to be moved is limited in one activation process of a GC task, and the compaction process is not performed in one continuous process, but is realized as an accumulation of short-time processes. Therefore, if another processing request occurs during compaction processing, the processing request will be kept waiting for at most one GO task processing time.
The response performance of the system hardly deteriorates.

It is assumed that the swap task and the GC task are not subject to allocation/release processing or save/recovery processing, and are always allocated to a predetermined location on the main memory.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 shows the hardware configuration according to the embodiment.

The hardware of this embodiment is a CPU that performs data processing.
(Processing device!, main memory 2 that stores programs and data executed by the CPU 1, hard disk (hereinafter abbreviated as HD) which is a secondary storage device that stores programs and data
3 and 70 tips 4, a keyboard 5 and a mouse 6 for inputting characters and functions, a display 7 for displaying processing results, etc., and a printer 8 for printing processing results, etc. It is comprised of an address converter 9 that converts a logical address, which is a memory address output by the CPU 1, into a physical address, which is a memory address used when accessing the main memory 2.

Figure 3 shows the configuration of the address converter 9.

The address converter 9 includes an address comparator 91, an address switch 92, an address adder 95, and a base register 94.

The address comparator 91 manually calculates the logical address and outputs a comparison result to determine whether the value of the input logical address is greater than or equal to the division point address preset in the address comparator 91.

The division point address is an address that indicates the boundary between the direct space and the window space in the logical address space.

The base register 94 is a register whose value is set by the CPU 1, and holds an addition offset value when performing address conversion.

The address adder 95 inputs the logical address and the value of the base register 94, and outputs the sum of the two.

The address switch 92 receives the logical address, the output of the address adder 95, and the output of the address comparator 91 as inputs, and according to the comparison result of the address comparator 91, if the value of the logical address is smaller than the division point address, the logical address The value of the address adder 93 is output as is, and if the value of the logical address is greater than or equal to the division point address, the output value of the address adder 93 is
Each of these is the output of the address switch 92.

FIG. 4 shows how the address converter 9 converts a logical address into a physical address.

The logical address space is divided into a direct space and a window space according to the division point address. If the logical address value is in the direct space, the logical address value and the physical address value are equal; if the logical address value is in the window space, the physical address value is equal to the logical address value in the base register 94. The value is the sum of the values. By converting the value of the base register 94, the correspondence between logical addresses and physical addresses in the window space can be dynamically changed.

FIG. 1 shows the software configuration of the system based on such a hardware configuration.

The software of this system includes an OS (operating system) 20, a swap task 30, a GCCjj page collection task 40, and a plurality of AP (application cage 1) tasks 50.

0820 realizes multitask processing of the swap task 30, GC rifuri 40, and AP task 50. Multitask processing is a widely known control method for combinator systems, so a detailed explanation will not be provided.

Swap task 30 swaps A between main memory 2 and ID3.
P task 50 is transferred to an AP whose capacity is larger than that of main memory 2.
Enables task 50 to be executed.

The GC task 40 moves the position where the AP task 50 is placed on the main memory 2 and stores more AP tasks on the main memory 2.
Allow task 50 to exist.

The AP task 50 performs data processing such as document editing processing and database processing required by the system. The AP task 50 does not always exist in the main memory 2, and if it is not being executed, it may be saved to the ID3, so such a task is called a non-resident task.

The location of the AP task 50 on the main memory, that is, the physical address, is not fixed and changes dynamically, but since 0820 changes the value of the base register 94 of the address translator 9, in the logical address space. , so that it always occupies a certain address. Explaining according to FIG. 4, the AP task 50 is arranged in a window space on the logical address space, and the logical start address of the AP task 50 is always equal to the division point address.

On the other hand, 0820, swap task 30, and GC task 40 are placed directly in the logical space, and their positions in the physical space do not change.

0820 is a system in which when operating a certain AP task 50, it is necessary to secure an area large enough to place the AP task 50 on the main memory 2.
The main memory 2 is divided into fixed-length pages 60 and managed.
The main memory allocation for P task 50 is page 60.
Perform this in the buttocks position.

The size of the AP task 50 varies depending on each AP task, and its maximum size is the size of the window space on the logical address space. In this way, in order to allocate memory for the variable capacity heron, the size of one page 60 is
A value smaller than the maximum AP task size (for example, 1/8
), one or more consecutive pages 60 are assigned to one AP task 50 for processing.

Since the number of pages 60 in the main memory 2 is finite, when different AP tasks 50 operate one after another, at a certain point, A
It becomes impossible to allocate page 600 to P task 50. In that case, save the AP task 50 that is not running at that time to ID3, release the page 600 assignment for that AP task 50, create a free space in the page 60, and move the AP task 50 that should be running. Assign.

) When the AP task 50 that was saved to ID5 is to be operated thereafter, the contents of the saved task are restored to the main memory 2 and operated. In that case as well, it is necessary to allocate the page 60 to the AP task 50, and if necessary, other AP tasks 50 are evacuated and then restored to the main memory 2. Such evacuation to the AP task 500HD3 is called swap-out processing, and restoration to the main memory 2 is called swap-in processing.

When such swap-in/out processing is repeated, pages 60 that are not allocated to the AP task are scattered on the main memory 2, and the total number of unused pages is larger than the amount required for the AP task 50. Despite the large number of pages K, a situation arises in which the size of the individual consecutive unused pages scattered about cannot be allocated because none of them are large enough to cover the required amount. This situation is said to be fragmentation occurring in the main memory 2.

Fragmentation is achieved by performing memory-to-memory copy on main memory 2.
This problem can be solved by moving the position of the task 50 and collecting unused pages in one place. The process of collecting unused pages in one place in this way is called compaction process.

Generally, the time required for compaction processing is proportional to the amount of allocated pages 60 in main memory 2, so the smaller the total amount of unused pages and the greater the need for compaction processing, the longer the compaction processing time will be. It has the property of being long.

In the present invention, by executing the compaction process when the system is idle, the load of the convax run process can be reduced, and the compaction process can be divided into short-time processes called unit transfer processes and executed, and arbitrary unit transfer processes can be performed. By making it possible to interrupt the compaction process at the end,
To perform compaction processing during idle time without reducing the responsiveness of the entire system.

Below, we will explain the specific data structure and program flow for implementing the processes described so far.

FIGS. 5 to 8 show data structures used in this embodiment.

In this embodiment, a task control block (hereinafter abbreviated as 'I'CB) 25 for managing tasks and a page control block (hereinafter abbreviated as 'I'CB) for managing a page 60 on the main memory 2 are used.
(abbreviated as B) 65, HD when saving tasks to HDS
Data of the evacuation control block (hereinafter abbreviated as E3CB) 35058i that manages E3CB is used.

FIG. 5 shows the configuration of the TCB 25.

The TCB 25 has a one-to-one correspondence with tasks, and not only the AP task 50 but also the swap task 30 and the GC task 40? Managed by CB25.

The TCB 25 includes a link pointer 2501, task priority 2502, task status 2503, and memory status 250.
4. Schedule time 2506, register save 2507
, task size 2508, task type 2509, PC
There are fields for a B pointer 2510 and an SCB pointer 2511.

Link pointer 2501 indicates the address of another TCB 25 and is used to create ready keys and various queues.

Task priority 2502 indicates the priority at which a task is executed.

The task status 2503 shows five types of status: paused, executable, and waiting.

A dormant state is a state in which a task exists but has not been started.A task in this state is not subject to execution.A task becomes executable when a start request is issued from another task. becomes.

The executable state is a state that is a target of task scheduling processing at 0320, and any task in this state is selected at 0820 and task processing is executed. A waiting state is a state in which a task makes various processing requests such as input/output requests and is waiting for their completion, and a task in a waiting state is not subject to execution.

The memory state 2504 is a state indicating the memory usage state of a task, and is a state independent of the task state 2505.

The task status 2504 has four types: evacuating, evacuating possible, evacuating prohibited, and movement prohibited.

The saving state means that the task does not exist in the main memory 2,
Indicates the status of being saved on the HDA.

The three states of evacuation possible, evacuation prohibited, and movement prohibited are all
This is the state when a task exists on the main memory 2, and the evacuation enabled state is a state in which the task can be saved to the HD 3 and the task can be moved on the main memory 2, and the evacuation is prohibited. The state is a state in which data can be saved to the HD 3 but cannot be moved on the main memory 2, and the movement prohibited state is a state in which neither evacuation nor movement can be performed.

A task is normally in an evacuation-enabled state, but becomes in an evacuation-prohibited state, for example, when it is acquiring system resources using a semaphore, etc. In such cases, it is necessary to prevent deadlocks from occurring. Therefore, saving tasks is prohibited.

In addition, the movement prohibited state occurs while the task is issuing an input/output request, and data transfer between the main memory 2 and the HD 3 etc. is performed by directly specifying the physical address without going through the CPU 1. Therefore, saving or moving tasks is prohibited during that time.

Schedule time 2506 is used to hold the latest time at which a task was selected for execution by 0820. The schedule time 2506 is used when the swap task 30 determines a task to be evacuated.

The register save 2507 is an area for storing the values of internal registers of the CPU 1, and is used to correctly suspend/restart task processing. When the execution of a certain task is interrupted to start execution of another task, the value of the internal register of CPU 1 at that point is saved in the register save 25.
07, and when the processing of that task is resumed, the value stored in the register save 2507 is saved to CP.
By resetting the internal register of U1, it is possible to restore the same task execution environment as at the time of interruption, and task processing is restarted correctly.

Task size 2508 is the number of pages 60 required to execute the task, and is used when allocating pages 60 to a task.

The task type 2509 indicates the type of memory usage of the task, and there are three types: resident, reusable, and non-reusable.

Resident is the type of the swap task 50 and GC task 40, and the AP task 50 has either a reusable type or a non-reusable type. A resident task has a fixed location in the main memory 2, is never saved, and always exists in the main memory 2. Reusable type and non-reusable type tasks are task types that can be saved to HD3, and these two types are collectively called non-resident types. The difference is that after a task finishes processing and moves from the executable state to the dormant state, when it is restarted and moves to the executable state, the memory image of the task that was in the executable state before 1 is restored again. The question is whether or not it can be used.A reusable type can be used, but a non-reusable type requires a new initial memory image of the task to be loaded onto the memory.

The PCB pointer 2510 stores the address of the PCB 25 indicating the page 60 allocated to the task.

When the value of PCB pointer 2510 is 00, it indicates that page 60 is not allocated to the task.

The SCB pointer 2511 is set when the task is being saved.
The address of 5CB35 indicating the area where the task is saved is stored. If the value of the SCB pointer 35 is 0, this indicates that the task has not been saved.

FIG. 6 shows the configuration of 5CB35.

A plurality of task save areas having the same size as the maximum task size are secured on the HD 3, and these task save areas are managed by the 5CB35.

5CB35 has HD position 3501 and evacuating task 3.
There are two fields: 502.

HD upper [3501 is the HDS of the task save area above]
The task 3502 that is being saved holds the address of the TCB 25 of the task that has been saved to the task save area. If no task has been saved in the task save area, the value of the task being saved 5502 is 0.

FIG. 7 shows the configuration of the VCPCB 65.

The PCB 65 includes a page connection state 6501, a page state 6502, a busy 7 lag 6503, and an update flag 6504.
, page type 6505, page length 6506, assigned task 6507, and empty page link 6508.

The PCB 65 has a two-to-two correspondence with the pages 60, and manages the relationship between pages and the relationship between pages and tasks.

A series of two or more consecutive pages is called a page series.
The connection state between pages is page connection state 6501iC
Therefore, it is managed.

The page connection state 6501 has four states: first, intermediate, final, and independent, and the page whose value in the page connection state 6501 is "first" is called the first page, an intermediate page,
The terms "final page" and "independent page" have the same meaning.

A page series is an independent page or 0 pages after the first page.
Refers to a series of pages that are lined up with one or more intermediate pages and one final page.

The state of a page series is managed by the state of its first page, that is, an independent page or a first page. Therefore, in the intermediate page and the final page, the contents of fields other than the page connection status 6501 in the PCB 65 are invalid and have no effect on the page series status.

The page status 6502 takes on five types of values: unused, in use, and reserved.

Unused indicates that the page is not assigned to a task, while used indicates that it is assigned to a task, and reserved indicates that the page is used for purposes other than storing tasks. Indicates that this is a page series that cannot be deallocated.

Page series are currently reserved for KO320 on that page.
This is the case when there is a resident task.

The busy 7 lag 6505 is unavailable if the busy flag 6503, which indicates whether the task assigned to the page set can use the page set, is true.
Busy7/l indicates available if false
The flag 6503 is true if the evacuation/recovery processing or compaction processing is being performed for the page series.
When these processes are completed, it becomes false.

The update flag 6504 becomes true when the task assigned to the page series is dispatched, and the contents of the page series are HD! Becomes false when SK is evacuated. Update flag 6
504 indicates whether there is a possibility that the contents of the page series have changed, and if true, there is a possibility that the contents of the page series in memory and the contents saved in HO2 do not match. If it is false, it means that the contents of the two match.

The page type 6505 indicates the type of the page series. Types of page series include system type and task type.

The system type is the type when the page series is directly used by the 0820 or a button such as a resident task, and the task type is the type when the page series is managed by the 0820 and used to load a non-resident task. It is a type.

When the page type 6505 is system type, the page status 6502 is always reserved. In addition, if the page type 6505 is a task type, the page state 6502
will be either unused or in use, and will not be reserved.

The page run length 6506 indicates the number of pages that make up the page run. For example, if there are three intermediate pages following the first page and then the last page, the length of the page run is 5. , 5 is set in the page length 6506 of the first page.

The first page of the next page series after a certain page series can be found by checking the page concatenation status '6501 of each page one after another, but by using the value of the series series length 6506, it can always be kept constant. The first page of the next page series can be found in the time .

The assigned task 6507 indicates the address of the TCB 25 of the task assigned to the page group.

If no task is assigned to the page series, the value of the assigned task 6507 is 0.

The empty page series link 6508 has meaning only when the page status 6502 of the page series is unused. Used to speed up searching for unused page sets.

FIG. 8 shows other data structures used in this embodiment.

The ready queue header 2601 indicates the address of the first TCB 25 of the ready queue, which is configured by connecting TCBs 25 in which the task state 2503 is executable using the link pointer 2501.

Empty page series header 2602 is page-shaped fi6502
PCB 65 which is unused is empty page link 650
P at the beginning of a list of empty pages connected by 8's
Indicates the address of CB65.

The idle time counter 2603 is used to measure the time in the idle state when there is no task to be executed, and is used to determine the timing when executing processing to return to the idle state. used.

The above is an explanation of the data structure used in this embodiment.

Next, the processing performed by this embodiment using the above data structure will be explained using an outflow diagram.

FIG. 9 is a flowchart of the task scheduling portion of the 0820 process.

0820 first, in step 2001, each TC in the ready queue connected to the ready queue header 2601.
The task priority level 2502 of B25 is checked, the task with the highest priority is selected, and the process proceeds to step 2002.

If the value of the ready queue header 2601 is 0, this indicates that there is no task in an executable state.
If the task can be selected, proceed to step 200.
2, the process advances to step 2005 and the selected TC
Check the value of the memory status 2504 of B25, and if it is being saved, proceed to step 2004; otherwise, proceed to step 2.
Proceed to 005.

In step 2004, the swap task 30 is started using the selected 'I'CB 25 as a parameter, and then the process returns to step 2001.

Swap task 50 is a task scheduled by 0820 and has the highest priority among all tasks, so swap task 50 is always scheduled in step 2001.
0 is selected and executed, and through the processing of the swap task 50, the swap-in processing of the task selected first is performed. The processing of the swap task 50 will be described later.

Now, step 20 when the task exists in memory
In the process of 05, the busy 7 lag 6503 of the page series assigned to the task is checked to determine whether the page series can be used by the task.

If the busy flag 6503 is true, it is not possible, and in that case, the process moves to step 2010, selects the task with the next highest priority from the ready queue, and performs step 2010.
002 and attempts to execute the task.

If the busy flag 6503 is false and the page set to which the task is assigned is usable, the process proceeds to step 2006, where preparations for starting the task are made.
6, scheduling 2506 in '1' CB25
Set the current time to , and then in step 2007,
PCB 650 update flag 65 assigned to the task
04 is true.

Next, in step 2008, a value is set in the address translator 90 pace register 94 and the window space mapping is changed so that the task to be executed is visible on the logical address space.

The value to be set in the pace register 94GC can be determined as follows.

The PCB pointer 2510 of the TCB 25 holds the address of the first page of the page series assigned to the task, but since the PCB 65 and page 60 have a one-to-one correspondence, the corresponding PCB address is used. It is possible to specify the number of pages out of n.

Since the length of each page is constant and a value unique to the system, the physical address of the first page can be found by the following formula: Physical address - Page length X(i-1). The value obtained by subtracting the division point address from the physical address of the page obtained in this way is
This is the value to be set in the base register 94.

When the switching of the logical address is completed, the task execution environment is restored in step 2009, the process of 0820 is completed, and task processing is restarted. To resume task processing, TCB
This is done by setting the contents of the register save 2507 of No. 25 in the internal register of the CPU 1.

Now, step 2 when there is no task to execute.
At step 020, first, the value of the idle time counter 2603 is set to 0.

Next, in step 2021, the idle time counter 26
The value of 03 is increased by 1, and it is determined whether it exceeds a preset value. If not, go to step 20 again.
The process returns to step 2021 to continue counting the time, and if the time has exceeded the time, the process proceeds to step 2022, where the GC task 40 is activated with the normal mode as a parameter, and compaction processing is performed.

The GC task 40 is a task with the next highest priority after the swap task 50, and its processing details will be described later.

Although the specific value to be set for measuring the time in step 2021 varies depending on the speed at which the CPU 1 executes the command, a value that is 300 to 2 seconds when converted to seconds is set.

The factors that cause the system to be in an idle state can be roughly divided into HD! 5. As in 70 Tsupi 4, there are cases where a response is being waited for for a peripheral device button that can be expected to return a response after a certain period of time, and as with keyboard 5,
Since a response is made in response to a user's operation, there are cases where the response is waited for with an indefinite response waiting time.

A characteristic of systems that perform interactive processing is that the waiting time for the latter response is overwhelmingly longer than the waiting time for the former response. One of the features of the invention is that in order to do so, it is necessary to distinguish between two types of waiting.
This is done in step 2021.

The response latency for HD5 is typically several tens of milliseconds, and the response latency for 70Tpi4 is 200 milliseconds.
J seconds or less. Therefore, if the idle state continues for 300 milliseconds, it can be determined that the device is waiting for a response from the user. On the other hand, if the user does not press any key for more than 2 seconds, it can be determined that the user has stopped inputting and is thinking. , it is possible to perform compaction processing while the user is interrupting input work.

Next, a flowchart of the processing of the swap task 30 is shown in FIG.

The swap task 50 is a task to be swapped in.
It is started with the L'CB address as a parameter, and first,
In step 3001, it is determined whether a set of pages of the required size can be allocated.

The size of the page series to be allocated is determined from the task size 2508 in the 'I' CB 25. The search for a page series to be allocated with 0 is performed in the following three stages.

First, the process is performed for pages whose page status 6502 is unused and whose g-flag is false, and then,
This is performed for page series in which the page status 6502 is in use and the update flag and busy7 lag are false. In either case, the search is performed with the first priority (first − fit +
) method, and the search is stopped when a page series (or a series of pages) larger than the desired size is found for the first time, and that page series is taken as the search result.

If assignment is possible, the process advances to step 3006, and the page series is assigned to the task. When allocating a set of pages that are currently in use, first, the assignments of tasks in the currently used block set are canceled to make them unused, and then the blocks are assigned to a new task.

To cancel task assignment of a page series, set the PCB pointer 2510 in the 'rCB25 of the task assigned to the page series to 0, and set the memory status 2503 to "Saving", and then
CB650 assigned task 6507 is in old state 65
This is done by setting 02 to unused.

In addition, to allocate a page series to a task, set the address of the page series PCB 65 in the PCB pointer 2510 in the TCB 25 of the task to be assigned, and change the memory state 250.
4 can be saved, then set the TCB address of the task to be assigned to the PCB 650 assigned task 6507,
This is done by setting the page status 6502 to be in use.

When allocating a page series, if the length of the page series is larger than the assigned task size, divide the page series into a page series of the same size as the task size and the remaining page series, and then divide the page series as needed. Assign a set of pages of size to a task. In addition, when setting the page status 6502 of a page series as unused, check the status of the page series before and after it, and if there are consecutive page series that are unused, merge those page series to create one unused page series. We also perform the task of organizing pages in their usage status.

When the assignment of the page series is completed, the process advances to step 3007, and a swap-in process is performed to restore the task contents to the assigned page series.

In the swap-in process, first, the busy flag 6505 of the page group assigned to the task to be swapped in is set, and then the contents of the task are read from the HD 3 onto the main memory 2. The upper position is determined from the HD position 5501 of the 8CB 55 indicated by the SCB pointer 2511 in the '1' CB 25, and the transfer destination location in the main memory 2 is determined from the value of the PCB pointer 2510 in the TCB 25.

When the transfer is completed, the task being saved is 1B4s5! 550
Set 2KO, SCB pointer 2 in 'I'CB25
511 is also set to 0 to release the allocation of the save area to the task. Subsequently, the busy flag 6505 of the PCB 65 is cleared, the swap-in process is ended, and the swap task 30 ends the process.

If the result of the determination in step 5001 is that the page series cannot be allocated, the process advances to step 3002, and it is determined whether there are any tasks that can be saved.

A task that can be saved is the task status 25 in the TCB 25.
03 can be saved and the busy flag 650 in the PCB 65 indicated by the PCB pointer 2510 in the TCB 25
3 is false, and the update flag 6504 is true. If such a task exists, step 300
5, if it does not exist, proceed to step 3005.

In step 3005, the GC task 40 is started using the priority mode as a parameter, and when the processing of the GC task 40 is completed, the process returns to step 3001 and attempts to allocate the page series again.

The processing of the GC task 40 will be described later.

In step 3003 when there is a task that can be saved, the schedule time 2506 in the TCB 25 is checked,
The task with the oldest scheduled time is determined as the task to be evacuated, and then the swap-out process for the selected task is performed in step 3004. When the swap-out process is completed, the process returns to step 3001 and the page series is allocated again. try.

Swap out processing is performed as follows.

First, search for 5CB35 in which the task 3502 being saved in 5CB35 is 0. If such an SCB does not exist, immediately terminate the swap-out process because there is no space in the save area on J(D3). If there is a vacant evacuation area, set the address of the 5CB35 in the SCB pointer 2511 in the TCB25, and set the TCB address of the task to be swapped out in the evacuation task 3502 in the 5CB35. Assign an evacuation area to a task.

Next, the busy 7 lag of the page series assigned to the task is set to true, and the page series contents are transferred from the main memory 2 to the HD 3.The transfer position is determined in the same way as any swap-in process. be able to.

When the transfer is completed, the PCB 650 update 7 lag 6504 is set to false, the busy 7 lag 6503 is also set to false, and the swap-out process is completed.

Figure 1 shows a conceptual diagram of unit transfer processing when performing compaction processing.

The unit transfer process includes slide transfer process and swap transfer process, and the contents of each process are shown in FIG. 11(a).
and shown in FIG. 11(b).

Before explaining the unit transfer process, first, the state of the page series to be subjected to the compaction transfer process will be explained.

The page series targeted for compaction processing are the page series whose page status 6502 is unused and the busy 7 lag 6505 is false, and the page series whose page status 6502 is in use and the busy flag 6503 is false. Memory status 250 in the TCB 25 of the task assigned to the page series
4 is a page series in a state other than movement prohibited. A page series in the latter state is called a movable page series.

Slide transfer processing is a transfer process performed between an unused page series and the movable page series immediately after it, and is a transfer process that is performed by copying the contents of the movable page series to the beginning of the unused page series in order. , moves the contents of the movable page series forward (in the direction of decreasing addresses) and moves the unused page series backward (in the direction of increasing addresses). In the slide transfer process, there are restrictions on the positional relationship between two sets of pages to be moved, but there are no restrictions on the length of the movement.

Swap transfer processing is a transfer process performed between an unused page series and a movable page series whose size is less than the unused page series, and the contents of the movable page series are transferred to the beginning of the unused page series. You can move the contents of a movable page series by copying them in order. At that time, if the length of the unused page series is longer than the length of the movable page series, the unused page series is divided into the movable page series and the unused page series, and the original movable page series is Change to an unused page series and change the position. In this case, the number of page series increases by one, and the fragmentation becomes more advanced. In swap transfer processing, there is a restriction on the length of two page series to be moved, but there is no restriction on the positional relationship.

By combining these unit transfer processes, the GC task 40 compacts the entire main memory 2.

FIG. 12 shows a flowchart of the processing of the GC task 40.

The GC task 40Vi is activated with the processing mode as a parameter. There are two processing modes: normal mode and priority mode. The normal mode is a mode activated when the system is in an idle state, and the priority mode is activated when the swap task 30 is performing swap-in processing. mode.

The GO task 40 first determines the startup mode in step 4001, and if it is the priority mode, the GO task 40 determines the startup mode in step 4002.
If the mode is normal mode, the process advances to step 4011.

In step 4002 in the priority mode, first, the page set at the highest address among the movable page sets is searched. The search results are determined in step 4005, and no items were found. In other words, if there is no movable page group, there is nothing that can be done, so
The processing of the GC task 40 ends.

If found, the process advances to step 4004 to check whether the found set of movable pages can be moved by swap transfer processing. A movable case is a case where an unused page series with a length longer than the found movable page series exists ahead of the found movable page series.

If it is possible to move, the process advances to step 4005, executes swap transfer processing once, and ends task processing. If there are a plurality of unused page sets to be swap transferred, the swap transfer process is performed on the unused page set at the lowest address among them.

If swap transfer cannot be performed, the process advances to step 4006 to determine whether slide transfer processing can be performed. The case where the slide transfer process can be performed is the case where the page series immediately before the found movable page series is an unused page series. If the transfer can be performed, the process advances to step 4007, performs slide transfer processing once, and ends the task processing.

If slide transfer cannot be performed, the process advances to step 4008, and searches for the page series existing at the highest address among the page series for which task allocation release processing can be performed for the page series. A page series for which task allocation release processing can be performed is one whose page status 6502 is in use, and whose page status 6502 and update flag 6503 and update flag 6 are in use.
504 is false.

The search results are judged in step 4009, and if found, the process advances to step 4010, where the process of releasing the allocated task for the page series is performed as explained in the swap out process of the swap task 30, and unused pages are released. As a result, the task processing ends.

If not found, proceed to step 4011.
Performs the same processing as when started in normal mode.

In step 4011 when started in normal mode,
First, a set of unused pages existing at the lowest address to which slide transfer can be performed is searched. An unused page series that can be transferred by slide refers to an unused page series whose immediately following page series is a movable page series.

The search result is determined in step 4012, and if found, the process proceeds to step 4013, performs slide transfer processing once, and ends the task processing.

If it is not found, there is nothing to do, so immediately end the task processing.

The above is the processing of GC task 40, and in normal mode,
The program attempts to eliminate fragmentation by moving low-level unused page series one after another toward high-level addresses, and in priority mode, creates a large unused page series at high-level addresses rather than eliminating fragmentation. Transfer processing is performed with this goal in mind. In either case, when the GC task 40 is activated once, only one unit transfer process is performed at most, and when compaction processing is performed, the CPU is not monopolized for a continuous long period of time.

According to this embodiment, since the compaction process of the main memory 2 is performed when the input from the user is interrupted for a certain period of time, the compaction process of the main memory 2 can be performed without increasing the overhead of the entire system. , the possibility that the AP task 50 exists on the main memory increases, and processing performance increases.

Furthermore, since the compaction process is divided into relatively small units called unit transfer processes, if the user inputs an input during the compaction process, the compaction process for the entire main memory 2 may not be completed. This has the effect that response processing to input can be performed by simply waiting until unit transfer processing is completed without having to wait for , and compaction processing can be performed with almost no change in system response speed.

〔Effect of the invention〕

As described above, according to the present invention, in a data processing device that performs interactive processing, main memory can be compacted without deteriorating responsiveness to user input or processing performance. This has the effect of reducing the number of times recovery processing is performed and improving the performance of the entire system.

In addition, the above functions can be achieved by using a simple address translation mechanism without using an expensive memory management mechanism such as a virtual memory system, which reduces system costs.
This has the effect of increasing cell performance.

[Brief explanation of the drawing]

Figure 1 is a software configuration diagram of an embodiment of the present invention, Figure 2 is a software configuration diagram of an embodiment of the present invention.
The figure is a hardware configuration diagram according to this embodiment, Figure 5 is an internal configuration diagram of an address converter, Figure 4 is an explanatory diagram showing a conversion method from a logical address to a physical address, and Figure 5 is a diagram of a task management block. 6 is a configuration diagram of the evacuation control block, FIG. 7 is a configuration diagram of the page control block, FIG. 8 is a configuration diagram of other data structures, and FIG. 9 is a flowchart of the processing of a part of O8. Figure 10 is a flowchart of swap task processing, Figure 11 is an explanatory diagram of page series unit transfer processing, and Figure 1
FIG. 2 is a flowchart of GC task processing.・1...C
PU 2... Main memory 3... Hard disk 4... Floppy 5... Keyboard 6... Mouse 7... Display 8... Printer 9... Address converter 20... Operating system 50...Swap Task 40...Kahesh Fist Collection Task 50...Application Task 60...Page. No. 0 No. 41 I ￥1 Sha Egg Punishment 1 r ￥1 PC et al. 5 8th mouth 7 wayoya 2601 1! 'F-f'Jti"26""tynt-+5sph;""' 50th "rc8+ Figure 6 q port 10th port 110

Claims (1)

[Claims] 1. Dynamically allocate a plurality of tasks in the main memory, and save/save the tasks between the main memory and the secondary storage device.
In a data processing device that performs recovery processing and performs multitasking processing, the compaction processing that consolidates the free areas in the main memory into one place is realized by repeating unit transfer processing, which is a short-time process that cannot be interrupted. , the above unit transfer process,
A task control method characterized in that the task control method is performed when the system of the data processing device is in an idle state. 2. Claim 1, wherein the unit transfer process is a process of moving one of the tasks within the main memory.
The task control method described. 3. The task control method according to claim 1 or 2, wherein the activation condition for the unit transfer process is that the system remains idle for a certain period of time or more. 4. The task control method according to claim 3, wherein the certain time is a value that can be determined to be a response waiting time for the user. 5. Dynamically allocate multiple tasks in the main memory and save/save the tasks between the main memory and the secondary storage device.
In a data processing device that performs recovery processing and multitask processing, a first table that manages the status of the task and the allocation status of the task in the main memory; and a first table that manages the usage status of the main memory. a second table; a third table that manages the save status of the task to the secondary storage device; and the first and second tables to consolidate the free area in the main memory into one location. To sum up, means for moving one of the tasks in the main memory, and using the first, second and third tables to allocate/release the task in the main memory. means for saving/restoring the task to the secondary storage device using the first, second, and third tables; means for activating the movement means when the task is not allocated in the main memory; and means for controlling the execution of the task, and when the task to be executed does not exist in the main memory, the allocation/release means and the evacuation/recovery means. 1. A task control system comprising: means for activating said moving means; and means for activating said moving means when a state in which there is no task to be executed continues for a certain period of time or more.