JPH01200451A - Internal data control method - Google Patents

Abstract

PURPOSE:To set the inhibition/release of send-out for each class by providing a protection flag that can be turned on and off at the application side. CONSTITUTION:At the application side, when a file to which the inhibition is desired for its send-out is designated at the inlet part of a certain process, the relevant information is stored into a file control table for the corresponding file via a file control module. Hereafter, when an access is given to the cluster of a page Pi belonging to said file, and therefore a protection flag Fp contained in the page map control information PCi to which said cluster corresponds is turned ON and the inhibition of send-out is set. While the send-out is permitted by turning off the flags Fp of all clusters belonging to the relevant file in an outlet process after the primary process is through.

Description

DETAILED DESCRIPTION OF THE INVENTION <Field of Industrial Application> The present invention relates to an internal data management method in a system capable of memory management using a virtual memory method, such as an automatic programming device.

<Prior Art> In an application program (simply referred to as an application) of an automatic programming device that automatically creates an NC program, it is necessary to handle a large amount of data that exceeds the capacity of an internal memory (real memory). For this reason,
In an automatic programming device that has a randomly accessible external memory (for example, a hard disk); a virtual memory method uses a part of the external memory as an extension of the internal memory (virtual memory area) to store data on the internal memory. Data is temporarily ejected to the virtual memory area, and data on the virtual memory area is restored onto the internal memory as necessary.

Then, data is purged and recovered in file units between the internal memory and external memory.

For example, if you run out of free space while adding data to a file in internal memory, the file will be removed using the following procedure. That is, (i) Closed files are ejected in order from the first file, and each time a file is evicted, it is checked whether the size is OK. At this time, your own files are excluded.

(If a sufficient size cannot be obtained with iil- or more, open files are evicted in order from the first file, and each time a file is evicted, it is checked whether the size is OK. At this time, the own file is excluded.

(2), (i) and (ii) if the file size is not large enough, the file itself is ejected. If the size is not obtained even in Gψ, (i) ~ field, it is assumed to be an overflow error.

However, with the conventional method of evicting files (:, the order in which files are evicted follows the originally defined order, it is not possible to specify the priority when executing the evicting, and
Since it is only possible to evict files in units of files, frequently used data is often also evicted, which increases external memory access and reduces processing efficiency. Also, even when using multiple files at the same time, if you pay attention to a particular moment, only a small amount of data in the file is actually needed, and the data that has been used once is likely to be used again in the near future. Although expensive, traditional methods do not allow portions of files to remain in internal memory, resulting in increased external memory access.

For this reason, the applicant introduced the concept of cluster,
Evicting data in cluster units instead of file units,
In addition to performing recovery, the information that manages each cluster on internal memory includes a mapping priority that indicates how difficult each cluster is to be evicted to external memory, and a mapping priority that indicates the difficulty of evicting clusters to external memory for each mapping priority. We have proposed an internal data management method in which a matrix is provided to indicate the difficulty of processing data, and when there are no free clusters in the internal memory, the first clusters in the pine pink matrix are flushed out to the external memory in order from the lowest priority cluster.

<Problems to be Solved by the Invention> By the way, depending on the application, there are cases where ``a certain file is intended for very high-speed access, so it is never wanted to be evicted.'' However, the proposed mapping priority can only express the difficulty of being kicked out, but it does not guarantee that it will not be kicked out, and there is a possibility that it will be kicked out.

Also, data removal between internal memory and external memory/
On the internal processing side that controls recovery, there are cases where you want to access two clusters alternately, but if you read one cluster into internal memory and then read another cluster into internal memory in the same way, the first clusters may be evicted. Therefore, it is necessary to check these and convert the cluster number and physical address every time, resulting in a large processing overhead.

From the above, the purpose of the present invention is to enable mapping to be evicted from the head of a matrix with a low mapping priority, and also to allow the application side and internal processing side to set prohibition/cancellation of evicting for each cluster, so that prohibited clusters are not evicted. An object of the present invention is to provide an internal data management method as described above.

Means for Solving the Problems> FIG. 1 is an explanatory diagram of an internal data management method according to the present invention.

ETM is external memory, INM is internal memory, PGA is page area in internal memory, VCT is management information area, VM
A is the virtual memory area, p, (o=.

~M) is a page, PC, Happy Shimatsubu management information, Fp
is a protect flag, and Fu is a working flag.

MPR is mapping priority.

Function> d-shimatsubu management information that inputs and outputs data between the internal memory INM and external memory ETM in fixed-length cluster units, and manages each page P1 (page length = cluster length) of the page area PGA. pc, and in the management information p and c (1) External memo of the cluster mapped for page P6! 0 represents the difficulty of being kicked out by J ETM
A mapping priority MPR of ~15, (1) a pointer NXP that indicates how difficult it is for a cluster to be evicted to the external memory ETM by a mapping priority chain saw, and (2) whether or not each cluster is to be evicted. When there is no free page in the page area PGA, the protect flag Fp and working flag F1 . Clusters where both l are off are found, and the clusters are stored in external memory (ETM
) to drive out.

<Embodiment> FIG. 1 is an explanatory diagram of an internal data management method according to the present invention.

ETM is an external memory such as a hard disk or floppy disk, and a part of it is the virtual memory area V.
It is MA. The virtual memory area VMA can exchange data with the internal memory INM in cluster units, and has multiple storage areas C0 to CN (cluster areas O to CN) of the cluster length.
It has a cluster area N).

INM is a real memory (internal memory) provided inside the system body, PGA is a page area, and VCT is a management information area. If the page area PGA is a large number of pages P0
~p+, (page 0 to page M), and each page stores data in units of clusters. Note that the page length=cluster length.

The management information area VCT includes (i) a pit map FAT indicating the free status of each page P1 to P of the page area PGA, and (1)) storage area in cluster units (cluster area) C0 to C0 of the virtual memory area VMA. A bit map CAT indicating the free status of CN, a page map PMAP having management information PC, (i=0 to M) corresponding to each page P0 to P□ of the jil page area PGA, and a page map PMAP of the Gvl virtual memory area VMA. Each cluster area C8~,
When setting a cluster map CMAP having management information CC, ~CCN corresponding to CN, and a mapping priority (0 to 15) representing the difficulty of expelling clusters in each M page, each value of the mapping priority ( A matrix indicating the difficulty of eviction of a cluster is formed for each matrix (0 to 15), and a mapping priority queue MPQ is provided to indicate the first page pointer PHDK (k = 1 to 15) and the last page pointer PTLK of each matrix. It is being

The management information PC1 provided corresponding to each page P, (i=o to M) of the page map PMAP includes (a) a flag for prohibiting the expulsion of clusters on page P; A protection flag Fp that can be turned on/off on the Mapping priority MPR indicating the difficulty of eviction of clusters in (d1 page P), and virtual memory area V
When the page is evicted to the MA, in the matrix that has the cluster number CLN that instructs which cluster area to eject and the same mapping priority as the tel page P1, the page P
a forward link NXP indicating the next page to be connected to page P1, and a reverse link PRP indicating the previous page to which this page P1 is connected in the matrix having the same mapping priority as (fll page). ing.

In addition, the management information CCJ provided corresponding to each cluster area C4 in the cluster map CMAP includes an instruction to which page in the page area PGA the data in units of clusters stored in the cluster area C4 is to be recovered. The page number PGN is included.

Note that the page number PGN in each management information CC1 indicates the position of the page map management information PC1 provided corresponding to the page P whose cluster is to be recovered.

Similarly, the first page pointer PHDK (K=1 to 15) and the last page pointer PTLK in the mapping priority queue MPQ both indicate the positions of page map management information corresponding to the first page and the last page of the priority matrix. In addition, each page map management information PC, forward link pointer NXP and reverse link pointer PR are shown.
P also indicates the position of the page map management information of the previous and next pages of the matrix.

Therefore, as shown in FIG.
The first page pointer PHD- of the mapping priority in PQ indicates the page map management information PC0 corresponding to the first page PP of the priority matrix, and the forward link pointer NXP of the page map management information PCp is assigned to the priority. Instructs the next page map management information PC in the matrix,
Forward link pointer N of page map management information PC0
XP is the next page map management information PC in the priority matrix
Instruct 1.

Further, the last page pointer PTL of the mapping priority in MPQ indicates the red map management information PC1 corresponding to the last page P of the priority matrix, and the reverse link pointer PRP of the page map management information PC is Similarly, the reverse link pointer PRP of the page map management information PC, points to the previous page map management information PC in the matrix with priority k. , instructs.

The mapping priority value (0 to 15) of each page P is specified by the application program for each cluster,
It is used to determine from which page data is to be evicted in clusters when free space is exhausted. Also, the page P in use is the page map management information PC,
It is always linked to some mapping priority matrix by the forward link pointer NXP and reverse link pointer PRP within.

Note that the first page in the same matrix is evicted, and when there are no more free pages, both the protect flag FP and working flag F of the first page indicated by the first page pointer P HDK of the lowest priority matrix are cleared. If it is off, the first page is deleted from the matrix and the cluster of the first page is ejected to the virtual memory area VNA. However, if any flag is on, the cluster on the first page is not evicted, and the protect flag FP and working flag F1 . Check whether l is on or off, and follow the same steps one after another until you find a page with both flags turned off. When you reach the end of the queue, set both flags from the first page for the next lowest priority matrix. Find the page where is off and kick out the cluster of that page.

On the other hand, if the mapping priority of the cluster in a predetermined page is specified by the application page 9, the page is linked to the end of the matrix of the specified mapping priority. Also, when a cluster of a predetermined page is accessed, or when a change in mapping priority is instructed, the link of the page is deleted from the corresponding matrix, and the link of the page is deleted from the corresponding matrix, and the link of the page is deleted from the corresponding matrix (if accessed, the same matrix). Link to the end of. Therefore, within a matrix, previously accessed pages are linked to the front of the matrix, and recently accessed pages are linked to the back of the matrix.

Next, an outline of operations for flag setting and expulsion will be explained.

When the application side specifies a file to be prohibited from being evicted at the entry point of a certain process, the file management module stores that information in a file management table (not shown) for the corresponding file. Thereafter, when a cluster of pages P belonging to that file is accessed, the protect flag Fp in the page map management information PC1 corresponding to that cluster (page) is turned on and eviction is prohibited.

On the other hand, eviction is made possible by turning off the protect flags of all clusters belonging to the file in the exit process after the process is completed.

b The internal processing side of the working flag loads the cluster that you want to prohibit eviction into internal memory during a certain process, and then
Working flag F1 in the page map management information corresponding to (page). Turn on l. As a result, the cluster always exists on the internal memory, so it is possible to use the physical address instead of the cluster number.

After the processing is completed, the working flag F of the cluster is turned off to enable eviction. However, there are cases where the protect flag Fp is on, but in this case, the ejection is still prohibited.

teacher) kicked out FIG. 3 is a flowchart of the eviction process.

When there are no free pages in the page area PGA, the first page of the matrix with the lowest mapping priority m (m=0) (if there is no first page, the first page of the next lowest matrix (m=1), and so on) pay attention (step 101).

Thereafter, it is checked whether the protect flag Fp in the page map management information corresponding to the page of interest is on (step 102).

If the protect flag FP is off, then check if the working flag F is on.Step 1
03) If it is off, the cluster of the page of interest can be evicted, so it is evicted to a predetermined cluster area of the virtual memory area VMA indicated by the cluster number CLN in the page map management information (step 104).

On the other hand, if either the protect flag Fp or the working flag F is on, the cluster of the page of interest is prohibited from being evicted, so check whether the page is the last in the matrix (step 105), and if it is not the last, the next evicted The ranking page is determined (step 106), and the processes from step 102 onwards are repeated.

In step 105, if the page of interest is the last page in the matrix, it is checked whether the matrix has the highest priority (the last matrix) (step 107); if it is the last page, the process is terminated as an error, and the If not, focus on the first page of the next matrix (step 108), and repeat the processes from step 102 onwards.

Specifically, the internal data management method according to the present invention described above is used, for example, in the following manner.

Aspect A: Two applications 1.2 are executed alternately, application 1 executes file 1 and application 2 executes file 1.
Now, suppose that file 2 is frequently accessed. (1) The protect flag for file 1 is turned on in the entrance processing of application 1 and turned off in the exit processing.
The protect flag is turned on and turned off in exit processing.

As a result, when one application is running, the files it uses can always be stored in memory, speeding up various processes.

Aspect B: When a part of a huge file is frequently searched within an application, the file itself is huge, so the mapping priority is set low in consideration of the impact on other files, and protection is performed at the beginning of a series of searches. Turn on the flag and turn it off at the end. As a result, clusters that have been searched twice remain in the internal memory, thereby reducing external memory accesses as a whole.

Aspect C When an application creates a new record, internal processing secures an area for record management information RCT, obtains a physical address, and then turns on the working flag of the cluster. After that, when you secure an area for actual data, write data, and set information in the record management area after writing, the cluster is guaranteed to be in the internal memory, so there is no need to recalculate the physical address. Information can be set quickly.

When internal data consisting of a record ID and real data for a record code is input from an application (creating a new record), the internal data management module converts the internal theta into variable-length actual data and fixed-length record management. The record management information RCT contains (1) the record ID of the actual data, its length, - the cluster (cluster number) of the page that stores the actual data, and the position information of the actual data in the 60 clusters. Let me have it,
Store actual data and record management information in separate pages (clusters).

<Effects of the Invention> According to the present invention, in addition to the mapping priority and access order matrix, a protect flag and a working flag are provided, so it is possible to set not only the difficulty of being evicted, but also the state in which the data will never be evicted. . As a result, the application side turns on the protect flag for files that are frequently accessed temporarily, or permanently, eliminating access to external memory, while the internal processing side uses clusters that are known to be reused. Processing can be sped up by turning on the working flag of the computer and storing and using its physical address.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram of an internal data management method according to the present invention, FIG. 2 is an explanatory diagram of a mapping priority matrix, and FIG. 3 is a flowchart of an expulsion process according to the present invention. ETM...external memory, INM...internal memory, P;...page, VMA...virtual memory area, MPQ...mapping priority queue, PCl...page map management information, Fp...protect flag, F...・Working Flag Patent Applicant Fanuc Co., Ltd. Agent Patent Attorney Chiki Saito Figure 2\, / -～-11111

Claims (4)

[Claims] (1) In an internal data management method in a system that has a randomly accessible external memory and uses the external memory as an extension of the internal memory, data is input between the internal memory and the external memory in fixed-length cluster units. In addition to outputting data, the information that manages each cluster on the internal memory includes a mapping priority that represents the difficulty of each cluster being evicted to external memory, and a mapping priority that represents the difficulty of evicting each cluster to external memory for each mapping priority. Each cluster has a matrix indicating its density and a flag indicating whether it is to be evicted or not, and if there are no free clusters in the internal memory, the flags are turned off in order from the first cluster in the matrix with the lowest mapping priority. 1. An internal data management method characterized by determining clusters of and expelling the clusters to an external memory. (2) There are two types of flags: a flag that can be turned on/off on the application side and a flag that can be turned on/off on the internal processing side, and clusters with both flags turned off are expelled to external memory. Internal data management method described in section. (3) Prepare management information corresponding to each cluster on the internal memory, and provide the mapping priority of the cluster corresponding to each management information, link data for forming a matrix of the priority, and the above two flags. An internal data management method according to claim 2, characterized in that the internal data management method includes: (4) Internal data according to claim 3, characterized in that the management information includes data indicating where in the external memory the cluster is to be ejected, and the data is referred to to eject the cluster to the external memory. Management method.