JPH0334015A - Input/output parallel processing method in quick sorting technique - Google Patents

Abstract

PURPOSE:To shorten the sorting process time by carrying out the input/output processes of the record of a part where the sorting process is completed and the record to be newly processed in parallel with a sorting process. CONSTITUTION:The input/output processes are repeated at every progress of a sorting process. Therefore since the new records to be sequentially sorted are already inputted to a record R1 area, a record R2 area... of a record storage area 4 when a 1st sorting process for all records stored in the area 4 with the initial input is completed, the 2nd sorting process can be immediately started. Then the time required to the sorting process can be shortened.

Description

[Detailed Description of the Invention] [Overview] During sorting processing, in particular, during sorting processing using the quick sort technique, the records of the part where the sorting processing has been completed are sequentially output, and the output is stored in the free area of the code next time. Concerning the parallel input/output processing method in the quick sort technique, in which sort processing is performed by sequentially inputting new records to be processed, the input/output of records for which sort processing has been completed and records to be newly processed The purpose is to shorten the sort processing time by executing processing in parallel with the sort processing.

Record Gunno il+! In a sort processing method that performs sort processing using the quinque technique on the records input to Ii and outputs the results, the records for which the sort processing has been completed are sequentially output from the record storage area during the above sort processing, and this output The processing configuration is such that new records to be processed next time are sequentially input into the record storage area that has become empty, and the processing is configured to sequentially input new records to be processed next time, sort processing, output processing of records for which sort processing has been completed, and new records to be processed next time. The record input process is configured to be processed in parallel.

[Industrial Application Field] The present invention sequentially outputs the records of the part where the sorting process has been completed during the sorting process, particularly the sorting process using the quick sort technique, and also stores the records in the free space of the outputted records next time. The present invention relates to a parallel input/output processing method in a quick sort technique in which sort processing is performed by sequentially inputting new records to be processed.

With the recent demand for faster computer systems, there is also a demand for faster sorting processing.

[Conventional technology]

In the conventional ascending order sorting processing method of sorting records stored in an external storage device in a predetermined order, for example, ascending order of numbers, two input processing, sorting processing, and output processing are sequentially and independently performed.

That is, FIG. 5 is a processing process explanatory diagram explaining the conventional processing method. Input processing is performed to input the record input into the input buffer 3 into the record storage area 4,
A CPU (not shown) performs sorting processing 1 in ascending order on the records stored in the record storage area 4, for example, by the quick sorting technique described above.

This quick sort technique involves dividing a group of records into a group of records with numbers smaller than the boundary value and a group of records with numbers greater than or equal to the boundary value, each of which is arbitrarily selected from a group of records. This is a technique for rearranging records in a predetermined order by repeating until a group consists of one record.

For the records stored in the record storage area 4, this quick sort technique is applied to the sort merge program and executed, the records sorted in ascending order are output to the work buffer 5, and are further transferred from the processing device 2 to the external storage device 6.
Output processing was being performed to output to.

[Problem to be solved by the invention]

As in the past, after waiting for the completion of all the sorting processes in which the above quick sort technique has been applied to the records stored in the record storage area 4, the records sorted in ascending order are transferred from the record storage area 4 to the work buffer 5. In the conventional sorting process, output processing is performed to output the data to the record storage area 4, and after waiting for this output processing to finish, input processing of a new record to be processed next from the input buffer 3 to the record storage area 4 is performed. Since this method is an independent type in which each process of input processing, sorting processing, and output processing is completely completed and then moving on to the next processing, it has the drawback that the sorting processing time is long.

The present invention executes the input/output processing of records for which sorting has been completed and records to be newly sorted in parallel with the sorting process, thereby making it possible to shorten the sorting processing time and parallelizing human output in the quick sorting technique. The purpose is to provide a processing method.

This will be explained in terms of the number of records stored in the external storage device l to be sorted.

(1) When the number of records stored in the external storage device l is larger than the number of records stored in the record storage area 4, the processing device 2 performs a sorting process, that is, as shown in FIG. 2(B), division into sizes is performed based on an arbitrarily selected boundary value Ll. Records having numbers equal to or higher than the boundary value 11 are stored in the record storage area 4. The remaining records having numbers smaller than the boundary value LL are divided into sizes based on the arbitrarily selected boundary value L2 (L2<LL), and the records having numbers equal to or higher than the boundary value (iL2) are Group ■ is stored in the record storage area 4. The record group in the range with large numbers divided in this way is always stored in the record storage area 4, and the boundary value is set for the records in the range of records in the range with small numbers. A quick sort technique is used to determine and divide the data, and the process is repeated until the quick sort limit is reached.

When the number of records to be divided becomes less than the limit value, a direct insertion method is performed in which the positions of adjacent records are swapped depending on the size of their numbers. There are 5 records in the record group ■ sorted in ascending order by this direct insertion method.
It is assumed that the sorting process has been completed and is output to the work buffer 5 as shown in FIG. 2(C).

Next, the processing device 2 takes out the record group {circle around (2)} which was divided last and is stored in the record storage area 4, and rearranges the records in this record group {circle around (2)} in ascending order by the direct insertion method. While sorting this record group,
The record group ■ for which the above sorting process has been completed is output to the work buffer 5, and records to be newly sorted are inputted from the input buffer 3 to the empty area ■ in the record storage area 4 as shown in FIG. 2(D). Ru. In the same way, during the next sorting process, the records for which this sorting process has been completed ■
is output to the work buffer 5, and a new record to be sorted is input from the input buffer 3 into the free area 2 in the record storage area 4.

Furthermore, the processing device 2 takes out the record group ■ which has been divided into the second to last and is stored in the record storage area 4, and performs quick sort on the records of this record group ■ until the quick sort limit value described above is reached. Divide the record group using the technique, and when it becomes less than the ξ value, rearrange the records within that record group using the direct insertion method. Output the records sorted by this direct insertion method to the work buffer 5. 1 Output to the work buffer 5 and input records to be newly sorted from the input buffer 3 into the empty record storage area 4. , such a sorting process is performed on the record group ■, and when the sorting process is completed, 1
Similar sorting processing is performed for the next record groups ■ and ■.

When the sorting process for all records stored in the record storage area 4 during initial input is completed and all records are output to the work buffer 5.

Since records to be subjected to the second sorting process are already stored in the record storage area 4, the processing device 2 immediately performs the second sorting process on all records stored in the record storage area 4. sort processing can be performed.

(n) When the number of records stored in the external storage device l is smaller than the number of records stored in the record storage area 4. In this case, at the initial input stage, the records are transferred from the external storage device l to the record storage area 4 via the input buffer 3. All records to be sorted have been input, so only the sorting and output processing in case N) are processed in parallel. Since the records for which the sorting process has been completed are sequentially output to the work buffer 5, the time required for the sorting process is shortened.

Although the above explanation is given for ascending order, the same can be done for descending order.

FIG. 3 is a record transfer explanatory diagram for explaining the record transfer state in the input/output process of the present invention, and FIG. 4 is a time chart of one embodiment thereof.

It is assumed that records to be subjected to the first sorting process have already been input into the record storage area 4 by initial input. The explanation will be as follows with reference to the time chart of FIG.

That is, when the sorting process is started for the records in the record storage area 4, the input covers 3-1.
When 1-2 is empty, records (a) and (b) stored in the external storage device 1 are transferred to the human cover sofa 3-1. Waiting for the completion of this transfer, records (c) and (d) stored in the external storage device 1 are transferred to the input buffer 3-2. For convenience of explanation, the input buffers 3-1 and 3-2 will be described as having a capacity that is filled with two records (a) and 2).

The quick sort described above for the records in record storage area 4 divides the records into small groups using the quick sort technique until the number reaches the limit value, and when the number falls below the limit value, the range is sorted in ascending order using the direct insertion method. Record storage area 4 where processing is performed and the sorting process is completed first
The record (i) in the area ■ is output to the work buffer 3-1. ■ in this empty record storage area 4
The record (a) of the input buffer 3-1 is input to the area. The second record in area 2 in the record storage area 4 for which the sorting process has been completed is output to the work buffer 3-1. The record (b) of the input buffer 3-1 is input into the area (2) in the now empty record storage area 4. Similarly, the third and fourth records in the 05■ area in the record storage area 4 for which sorting processing has been completed (ha)
.

(in) are sequentially output to the work buffer 5-2.

For convenience of explanation, this work buffer 5-1.5-2 is also explained as having a capacity that is filled with two records.

Immediately after the record (b) is output from the input buffer 3-1 to the record storage area 4, or at an appropriate time, a new record to be sorted is sent from the external storage device 1 to the input buffer 3-1. (e) and (f) are inputted, and when records (ro) are inputted from the record storage area 4 to the work bar 5-1 and the work bar sofa 5-1 becomes full, immediately or at an appropriate time. , the work buffer 5-1 is empty after outputting records (i) and (ro) to the external storage device 6, so it is the fifth buffer.

Upon completion of the sixth sorting process, the records (A) and (F) in the ■ and [F] areas in the record storage area 4 are sequentially output to the work buffer 5-1. Also, records (E) and (E) are sequentially input from the input buffer 3-1 in the ■ and [F] areas in the record storage area 4 which have become vacant due to this output. The other input buffer 3-2 and work buffer 5-2 are also input buffer 3-1 and work buffer 5-2.
1 and the same operation is performed alternately, so that continuity of input/output processing is maintained.

In this way, each time the sorting process progresses, the input processing and output processing are repeated, and when the first sorting process for all records stored in the record storage area 4 due to initial input is completed, the records are stored in the record storage area 4. A new record expressed in katakana has already been stored, and the second sorting process is started immediately.

〔Effect of the invention〕

As explained above, according to the two aspects of the present invention, the input/output processing is performed in parallel during the sorting process, so the sorting process time is shortened, and the overall processing time is also shortened.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of record-to-record transfer for explaining the present invention in detail, and FIG. 2 (A) to D)! Fig. 3 is an explanatory diagram of disassembly processing of an embodiment of the invention; Fig. 3 is an explanatory diagram of record transfer to explain the state of record transfer in the input/output processing of the invention; Fig. 4 is a time chart of the embodiment; Fig. 5 is a conventional diagram. Fig. 11 is a processing progress explanatory diagram illustrating a sorting processing method. In the figure, l is an external storage device, 2 is a processing device, and 3°3-1.
3-2 is an input buffer, 4 is a record storage area, 5.5
-1.5-2 indicates that the work buffer is an external storage device.

Claims (1)

[Claims] Records stored in an external storage device (1) are input to an input buffer (3) in a processing device (2), and records in the input buffer (3) are input to a record storage area (4).
), the records input to this record storage area (4) are sorted using the quick sort technique, and the records sorted in a predetermined order are transferred from the record storage area (4) to the work buffer (5). ) via an external storage device (
In the sort processing method that outputs to 6), during the above sorting process, the records of the part where the sorting process has been completed are sequentially output from the record storage area (4) to the work buffer (5), and the records of the part where this sorting process has been completed are The processing configuration is such that the records are sequentially input from the input buffer (3) to the record storage area (4) which becomes vacant by outputting the record to the work buffer (5), and the sorting process and the sorting process are completed. 1. A method for parallel input/output processing in a quick sort technique, characterized in that output processing of a record of a previously processed portion and input processing of a new record to be processed next time are processed in parallel.