JPH07104814B2 - Invalid data detection method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention detects that a structure (data cell) allocated to a predetermined area of a storage device is no longer referenced and detects invalid data for effective use of memory. Regarding the method.

[0002]

2. Description of the Related Art An information processing apparatus is provided with a storage device for storing data for arithmetic processing. When writing data to this storage device, the data is allocated and written into an empty area of the storage device while dividing the data into appropriate units. Such a region is called a structure. For example, a program written in a single assignment language is suitable for parallel processing.
The structure generated during the execution of this program may be referred to many times by each processing that proceeds at the same time. Each structure allocated to the storage device is referenced by a write pointer and a read pointer. When proceeding with such processing, it is necessary to immediately detect a structure that is no longer referenced and incorporate it into a new memory resource so that the resource can be effectively used. Therefore, a method for promptly detecting that the structure is no longer referenced is introduced.

FIG. 2 is a diagram for explaining the relationship between a conventional pointer and a structure. The structure 3 allocated to a predetermined area of the storage device is referred to by an arbitrary number of pointers 1 as shown in this figure. (A) shows two pointers 1
Shows an example in which one refers to one structure 3. Each pointer 1 is provided with attribute information 2 shown in FIG. This attribute information is 1-bit information indicating an on or off state, and is information used in combination with the attribute information of the structure 3 described later to detect that the structure 3 is not referred to. Is. Each pointer 1 is used for any purpose such as reading or writing. Here, as shown in (a), first, when two pointers 1 are set for one structure 3,
The attribute information 2 is turned off. Further, as shown in (b), when three or more pointers 1 are set for one structure 3, the attribute information 2 is turned on.

FIG. 3 is a diagram for explaining a conventional pointer copy operation. The pointer 1 as described above can be increased by copying. That is, when there are one or more pointers, one of them can be copied to provide an arbitrary number of new pointers. Each pointer is 1
It is erased by writing and reading a number of times, and thus the pointer is incremented in this way when the same structure is referenced many times. First, as shown in FIG. 3A, when one pointer 1 is set for one structure 3, its attribute information 2 is off. However, when the pointer 1 is copied and increased to two, the attribute information 2 of the two increased pointers 1 is switched on as shown in the figure. This is also a device for detecting that the structure is no longer referenced, as will be described later. (B) shows a case where the attribute information of the original pointer is on. For example, as shown in FIG. 2B, when the structure 1 is referenced several times by the pointer 1 in which the attribute information 2 is on, the attribute information 2 is displayed as shown in FIG.
Only one pointer 1 with is on remains. When the pointer 1 is copied and increased to two in such a state, the attribute information 2 remains on.

FIG. 4 is an explanatory diagram of attribute information of a structure in the conventional case. In order to detect that the structure is no longer referenced, the attribute information 4 is also provided on the structure 3 side in this conventional example. Then, as shown in FIG. 4A, when the structure 1 is referenced by the pointer 1 whose attribute information 2 is off, the attribute information 4 of the structure 3 is turned off. Further, as shown in (b), when the structure 3 is referenced using the pointer 1 whose attribute information 2 is on, the attribute information 4 of the structure 3 is turned on. In addition, in the following description, each structure is illustrated by an expression as shown in (c).

FIG. 5 shows an example of a conventional invalid data detection method. In the conventional method shown in this figure, the attribute information 4 of the structure 3 read by the pointer 1 whose attribute information 2 is off is read.
Is off, it is determined that the structure 3 is no longer referenced. On the contrary, when the structure 3 may still be referred to, the pointer 1 or the structure 3
Any of the attribute information 2 to 4 is turned on. Invalid data is detected based on such a criterion.

That is, as shown in FIG. 5A, it is assumed that, for example, two pointers 1 whose attribute information 2 is off refer to one structure 3. In this case, when the structure 3 is referenced by one pointer 1, the pointer 1
Disappears, and only one pointer 1 is set for the structure 3 as shown in FIG. here,
As a result of referring to the structure 3 by the pointer 1 in which the attribute information 2 is off in (a), the attribute information 4 of the structure 3 is off as shown in (b). Therefore, FIG.
When the structure 1 is referred to by the pointer 1 in the state shown in FIG. 3B, the pointer 1 whose attribute information 2 is off is referred to the structure 3 whose attribute information 4 is off. As a result, as shown in (c), it is immediately detected that the structure 3 is no longer referenced after being referenced by the last pointer 1.

FIG. 6 shows an explanatory view in the case of detecting that the structure can still be referred to in the conventional case. Figure 6
As shown in (a), consider a case where, for example, there are two pointers 1 whose attribute information 2 is off. Here, as shown in (b), it is assumed that one pointer 1 is copied and the total number is increased to three. In this case, the attribute information 2 of the pointer 1 copied and increased is turned on. Therefore, one piece of attribute information 2 of the three pointers 1 is off, and the other two pieces of attribute information 2 are on.

Next, as shown in (c), the structure 3 is referenced once by the pointer 1 whose attribute information 2 is off, and when that pointer 1 disappears, two pointers 1 whose attribute information 2 is on are displayed. Remain. At this time, the attribute information 4 of the structure 3 is turned off. In such a state, no matter which pointer 1 is used to refer to the structure 3 thereafter, a criterion for determining that the structure 3 is no longer referred to,
That is, neither the pointer nor the structure satisfies the condition that the value of the attribute information is off.

(D) shows the state after referring to the structure 3 by using the pointer 1 with the attribute information 2 turned on once in the state of (b). In this case, in (d), the attribute information 2
There is one pointer 1 for which is off and one pointer 1 for which attribute information 2 is on. Even in such a case, even if any of the pointers 1 is used to refer to the structure 3 thereafter, the condition that the attribute information of both is off is not satisfied. Therefore, conventionally, a method of detecting that the structure is no longer referenced has been adopted only in the case shown in FIG.

[0011]

By the way, according to the conventional method as described above, even when the total number of pointers is three or more, or even when the total number of pointers is two, as shown in FIG. When the state shown in () is reached, it becomes impossible to detect that the structure is no longer referenced. Therefore, if you execute a program that generates many reference pointers to the same structure, it is necessary to detect the structure that is no longer referenced by another method and recover the resources. I will end up. However, the conventional method as described above has high utility value as a method for quickly collecting resources with a small overhead, and further improvement is desired. The present invention has been made in view of the above points, and provides an invalid data detection method capable of immediately detecting that a structure is no longer referenced by adding certain information to a pointer. The purpose is.

[0012]

According to the method for detecting invalid data of the present invention, it is detected that a structure allocated to a predetermined area of a storage device is no longer referred to, and the predetermined area of the storage device is detected. It is used for releasing, and at the time of allocating the structure, a write pointer and a necessary number of read pointers are set, and there is only one write pointer per one structure, There can be any number of read pointers for one structure, and the number of read pointers can be increased by copying.
It has three types of attribute information meaning GLE, DOUBLE, or MANY, and the structure is SINGLE or D.
If there is two types of attribute information meaning OUBLE and one read pointer is generated with the structure, the attribute information of the read pointer is set to SINGLE, and the read pointer is set to 2 with the structure. When this is generated, the attribute information of the read pointer is set to D
If the read pointer is set to OUBLE and three or more read pointers are generated together with the structure, the attribute information of the read pointer is set to MANY and the attribute information is SING.
When the read pointers set in LE are increased to two by copying, the attribute information of each read pointer is set to DOU.
When the number of read pointers whose attribute information is set to DOUBLE is increased to three or more by changing to BLE, the attribute information of each read pointer is changed to MANY,
When the structure is written, the attribute information of the structure is set to SINGLE, and when the structure is read by the read pointer having the attribute information set to SINGLE, the structure is no longer referred to. Detect
When the structure is read by the read pointer in which the attribute information is set to DOUBLE, if the structure attribute information is SINGLE, it is changed to DOUBLE and the attribute information is set to D.
When a structure whose attribute information is DOUBLE is read by the read pointer set to OUBLE, the fact that the structure is no longer referenced is detected.

[0013]

In this method, in order to identify the write pointer and the three types of read pointers, WRITE, SIN
2-bit attribute information of GLE, DOUBLE, and MANY is given. In addition, SINGLE or D for the structure
Gives 1-bit attribute information of OUBLE. Then, the attribute information of the structure is set to SINGLE immediately after writing the data, and when the attribute information is read by the read pointer of DOUBLE, the attribute information of the structure is changed from SINGLE to D.
Changed to OUBLE. The condition for detecting that the structure is no longer referenced is that the attribute information is read from the read pointer whose attribute information is set to SINGLE and that the structure whose attribute information is DOUBLE is read from the attribute information read pointer that is set to DOUBLE. When there is. As a result, even when there are three write pointers and three read pointers in total, it is possible to immediately detect that the structure is no longer referenced.

[0014]

The present invention will be described in detail below with reference to the embodiments shown in the drawings. FIG. 1 is an explanatory diagram of an invalid data detection method of the present invention. First, in the present invention, each pointer is clearly distinguished into a write pointer and a read pointer, and attribute information having three kinds of contents is added to the read pointer. FIG. 1A shows the definition of the attribute information. Attribute information 9 is provided in the read pointer 8 provided to refer to the structure 7. This attribute information 9 is SINGLE, D
There are three types, OUBLE and MANY. This SINGL
The attribute information of E is set when one read pointer 8 is generated together with the structure 7. Also, DOUBLE
The attribute information of is set when two read pointers 8 are generated together with the structure 7. Further, the MANY attribute information is set when three or more read pointers 8 are generated together with the structure 7. In order to identify this, in the figure, the attribute information of SINGLE is indicated by a white circle, the attribute information of DOUBLE is indicated by a black circle, and the attribute information of MANY is indicated by an X mark. The attribute information is represented by 2-bit data, for example, SINGLE is “01”, DOUB.
LE is represented as “10” and MANY is represented as “11”. In addition, when the attribute information is "00", it indicates a read pointer.

On the other hand, the structure 7 is also provided with similar attribute information 10. The attribute information 9 of the structure 7 will be described in detail later, but when data is written, S
INGLE and when the attribute information is read by the pointer of DOUBLE, SINGLE to DOUBL
It shall be changed to E. 1 (b) and (c)
Shows an example of a structure that is detected as no longer referenced by the present invention. Before describing this subject, first, the respective pointers, attribute information, and the structure of the structure will be specifically described with reference to FIGS.

FIG. 7 is an explanatory view of the write / read pointer according to the present invention. As shown in (a) of the figure, in the present invention, it is assumed that only one write pointer 6 exists for one structure 7. The write pointer 6 is clearly distinguished from the read pointer described above by its attribute information. Since the attribute information of the write pointer 6 is always WRITE, the illustration of the attribute information is omitted in the following figures. In addition, FIG.
The read pointer 8 for referencing the structure 7 can be increased to any number by copying, as shown in FIG. As shown in (c), a double-lined arrow is used to refer to the write pointer 6 and is shown separately from the case of reference by the read pointer 8.

FIG. 8 shows a pointer attribute information explanatory diagram according to the present invention. As shown in FIG. 8A, in order to write data to the structure 7 and refer to it, basically one write pointer 6 and one read pointer 8 are provided as shown in this figure. When only one read pointer 8 is generated in this way, the attribute information 9 is SINGLE. On the other hand, as shown in FIG. 8B, when two read pointers 8 are simultaneously generated, the attribute information 9 is set to DOUBLE. Further, as shown in (c), when three or more read pointers 8 are generated at the same time, the attribute information 9 is set to MANY. In (d), the shapes of the write pointer 6 and the read pointer 8 and their attribute information used in the following description are defined.

FIG. 9 is a diagram for explaining the pointer copy operation according to the present invention. As shown in FIG. 9A, when there is one read pointer 8 for the structure 7 and the attribute information 9 is SINGLE, the read pointer 8 is copied into two. There is. In such a case, the attribute information 9 of the read pointer 8 is SING.
Changed from LE to DOUBLE. Also, FIG. 9 (b)
As shown in, there is a case where the read pointer 8 which is initially provided only once is increased to three or more by copying. In this case, the attribute information 9 of the read pointer 8 is SI
Changed from NGLE to MANY.

Further, as shown in (c), one structure 7
However, there is a case where there is one read pointer 8 and the attribute information is DOUBLE. This occurs after the structure 7 is referenced by using one of the two read pointers 8 of DOUBLE as the attribute information 9, as shown in FIG. 9A. In the case of (c), when the read pointer 8 is increased by copying, as shown in the figure,
The attribute information 9 of the read pointer 8 is changed to MANY. Further, as shown in (d), when the read pointer 8 of MANY for the attribute information 9 is copied and increased, the attribute information is set to MANY.

FIG. 10 shows a pattern explanatory diagram of a read pointer that points to one structure. When implementing the present invention, focusing on the read pointer, there are patterns (a) to (d) as shown in the figure in order to refer to one structure. First, in (a), the attribute information 9 is SINGL.
This is a case where only one read pointer 8 of E exists.
Further, (b) is a case where there are only two read pointers 8 whose attribute information 9 is DOUBLE. In the pattern (c), the read pointer 8 whose attribute information 9 is DOUBLE
And two or more read pointers 8 whose attribute information 9 is MANY
Is present. In (d), the attribute information 9 is MAN
This is a case where there are four or more Y read pointers 8.
In practicing the invention, the read pointer 8 can take any of these patterns.

FIG. 11 is an explanatory view of attribute information of a structure according to the present invention. As shown in FIG. 11A, in the case of implementing the present invention, the attribute information 10 of the structure 7 becomes SINGLE when the data is written in the structure 7 by the write pointer 6. Further, as shown in (b), when the attribute information 9 refers to the structure 7 by the DOUBLE read pointer 8, the attribute information 10 is SINGL.
Changed from E to DOUBLE. As a result, the fact that the attribute information 10 of the structure 7 is SINGLE means that the read operation has not yet been performed for that structure, or that the read operation is performed once or more via the read pointer whose attribute information is MANY. It is either being done. The attribute information of the structure does not change when the attribute information is referred to by the read pointer of MANY.

The attribute information 10 of the structure 7 is DOUB.
Being LE means that a read operation via the read pointer 8 whose attribute information is DOUBLE is performed only once, or a read operation via the read pointer 8 whose attribute information is DOUBLE is performed once,
Further, it is a case where the read operation via the read pointer whose attribute information is MANY is performed once or more.

By setting the attribute information and the like as described above, it can be detected in the present invention that the structure is no longer referenced in the following cases. Returning to FIG. 1, the read pointer 8 whose attribute information 9 is SINGLE is 1 in FIG.
An example of the present structure 7 is shown. This structure 7
Causes the attribute information 10 to be SINGLE. In the present invention, the attribute information 9 is S
When the attribute information 10 refers to the SINGLE structure 7 by the read pointer 8 set to INGLE, it is detected that the structure is no longer referenced.

On the other hand, as shown in (c), first, consider a case where there are two read pointers 8 whose attribute information 9 is DOUBLE. At this time, the attribute information 10 of the structure 7 is SIN.
It is GLE. Here, when the structure 7 is referenced by one read pointer 8, the read pointer 8 disappears and the read pointer 8 whose attribute information 9 is DOUBLE is 1
Only books remain. Further, the attribute information 10 of the structure 7 is changed from SINGLE to DOUBLE as already described with reference to FIG. After that, the attribute information 9 is DO
The attribute information 10 is set to D by the read pointer 8 of UBLE.
When the OUBLE structure 7 is referenced, it detects that the structure 7 is no longer referenced. That is, the method of the present invention can detect that the structure is no longer referenced in the two cases shown in FIGS. 1 (b) and 1 (c). Therefore, in the case shown in (c), the number of patterns in which invalid data can be detected increases as compared with the conventional case.

FIG. 12 is a diagram for explaining a concrete example of the invalid data detection method according to the present invention. FIG. 12A shows the process from the time when the structure of the case shown in FIG. 1B was created. As shown in this figure, when the structure 7 is generated,
In this example, one write pointer 6 and one read pointer 8 are generated. Therefore, the attribute information 9 of the read pointer 8 in this case is SINGLE. Here, when data is written in the structure 7 using the write pointer 6, the write pointer 6 disappears and one read pointer 8 remains. This attribute information 9 is SINGLE
Has become. The attribute information 10 of the structure 7 is SINGLE by writing the data. Therefore, when the attribute information 9 or 10 is SINGLE and the read pointer 8 subsequently refers to the structure 7,
The detection conditions of the present invention described above are satisfied. This detects that the structure 7 is not subsequently referenced.
It should be noted that this case is equivalent to the conventional method shown in FIG. 2A, and there is no special improvement in this point compared with the conventional method.

On the other hand, as shown in FIG. 12B, it is assumed that one write pointer 6 and one read pointer 8 are generated for the initial structure 7. At this time, the attribute information 9 of the read pointer 8 is SINGLE. However, it is assumed that the read pointer 8 is subsequently copied and increased to two. In this case, the attribute information 9 of the read pointer 8 becomes DOUBLE. After that, when the data is written in the structure 7 using the write pointer 6, the attribute information 10 of the structure 7 becomes SINGLE.

When the structure 7 is referred to by using one of the two read pointers 8, the attribute information 9 is D
One OUBLE read pointer 8 remains. Further, the attribute information 10 of the structure 7 is changed from SINGLE to DOUBLE. In such a state, when the structure 7 is subsequently referred to by the read pointer 8, it is detected that the detection condition according to the present invention described above is satisfied and the structure 7 is not subsequently referred to. Therefore, in such a pattern, it is impossible to detect that the structure is no longer referred to, but the present invention makes it possible.

FIG. 13 is an explanatory view of an example to which the method of the present invention cannot be applied. The method of the present invention cannot be applied to the following cases. In this example, first the write pointer 6 and the two read pointers 8 are added to the structure 7.
Is being generated. In this case, the attribute information 9 is DOUBL
It is said to be E. However, after that, one of the read pointers 8 was copied, and the total number of read pointers 8 was three. In this case, the read pointer 8 used for copying and the read pointer 8 increased by copying are
In each case, the attribute information 9 is set to MANY, and when data is written in the structure 7 using the write pointer 6 in this state, the attribute information 10 of the structure 7 becomes SINGLE. Furthermore, the attribute information in the read pointer 8 is MANY
When the read pointer is used for reference, two read pointers 8 remain as shown in the figure, and one of the attribute information 9 is DOUBLE and one is MANY. Then, when the structure 7 is referenced by using the read pointer 8 of DOUBLE for the attribute information 9 next, one read pointer 8 for which the attribute information 9 is MANY remains at the end. In this case, even if the structure 7 is referred to by using the remaining one read pointer 8, the above condition is not satisfied.

Therefore, when the read pointer whose attribute information is DOUBLE is further increased due to copying or the like, the number of the remaining read pointers cannot be clearly expressed by the attribute information, and finally the last read pointer remains. Even if a structure reference is made by, it is not possible to distinguish whether this is the last. According to the method of the present invention, it becomes impossible to detect that the structure is no longer referenced when the read pointers having the attribute information of many are mixed.

The present invention is not limited to the above embodiments. In the above embodiment, SING is used to distinguish the pointers.
Although expressions such as LE, DOUBLE, and MANY are used, this expression is not necessarily limited to this and may be freely set. In short, it is significant to distinguish whether there are two or more read pointers by such attribute information. Further, although the expression such as SINGLE or DOUBLE is used for both the read pointer and the attribute information of the structure, a completely different expression may be used for these as well.

[0031]

According to the invalid data detection method of the present invention described above, when a structure is generated, it is possible to distinguish between a write pointer and a read pointer which are generated at the same time as a write pointer or a read pointer. , SINGLE, DOUBLE, MAN indicating the number of read pointers generated and the copy state of the read pointers thereafter.
Attribute information such as Y is added, and the state immediately after writing is performed on the structure side and the attribute information
Since the attribute information having the meaning of SINGLE or DOUBLE that distinguishes the case of reading by the read pointer set in LE is provided, the attribute information is combined with the SINGL attribute information by a combination of these attribute information.
When the structure is read by the read pointer set to E and when the structure of which the attribute information is DOUBLE is read by the read pointer whose attribute information is set to DOUBLE, the structure is no longer referred to. Can be detected. As a result, by adding a very small amount of information to the pointer, a structure that is no longer referenced is immediately detected, and the structure allocated to a prescribed area of the storage device is released to enable quick and effective use of memory. Can be planned.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an invalid data detection method according to the present invention.

FIG. 2 is a diagram for explaining the relationship between a conventional pointer and a structure.

FIG. 3 is an explanatory diagram of a conventional pointer copy operation.

FIG. 4 is an explanatory diagram of attribute information of a conventional structure.

FIG. 5 is a diagram illustrating an example of a conventional invalid data detection method.

FIG. 6 is a diagram illustrating detection of “can still be referred to” in the related art.

FIG. 7 is an explanatory diagram of a write / read pointer according to the present invention.

FIG. 8 is an explanatory diagram of attribute information of a pointer according to the present invention.

FIG. 9 is an explanatory diagram of a pointer copy operation according to the present invention.

FIG. 10 is a pattern explanatory diagram of a read pointer that points to one structure.

FIG. 11 is an explanatory diagram of attribute information of a structure according to the present invention.

FIG. 12 is a diagram illustrating a specific example of the invalid data detection method according to the present invention.

FIG. 13 is an explanatory diagram of an example to which the method of the present invention cannot be applied.

[Explanation of symbols]

 7 structure 8 read pointer 9 attribute information of pointer 10 attribute information of structure

Claims (1)

[Claims] 1. A structure used for releasing a predetermined area of a storage device by detecting that a structure allocated to the predetermined area of the storage device is no longer referred to, At the time of allocating the structure, a write pointer and a necessary number of read pointers are set, and there is only one write pointer per structure, and the read pointer is arbitrary for one structure. The read pointer has three types of attribute information meaning SINGLE, DOUBLE, or MANY, and the structure has two types of attribute information meaning SINGLE or DOUBLE. If one read pointer is generated together with the structure, the attribute information of the read pointer is SINGL.
When the read pointer is set to E and two read pointers are generated together with the structure, the attribute information of the read pointer is DOUBL.
If the read pointer is set to E and three or more read pointers are generated together with the structure, the attribute information of the read pointer is set to MAN.
If the number of read pointers set to Y and the attribute information set to SINGLE increases to two by copying, the attribute information of each read pointer is changed to DOUBLE, and the read pointer set to DOUBLE attribute information is copied. When the number is increased to 3 or more, the attribute information of each read pointer is changed to MANY, and when the structure is written, the attribute information of the structure is set to SINGLE and the attribute information is set to SINGLE. When the structure is read by the read pointer, it is detected that the structure is no longer referenced, and when the structure is read by the read pointer whose attribute information is set to DOUBLE, the structure attribute is read. If the information is SINGLE, change it to DOUBLE and set the attribute information to DOUBLE. Invalid data detection method attribute information when a reading of the structure of DOUBLE is characterized in that its structure to detect that it is no longer referenced by.