JPS6116326A - Sort processor - Google Patents

Abstract

PURPOSE:To perform effectively the sort processing using plural keys by outputting data in the input order if these data are equal to each other. CONSTITUTION:The first and the second memories MU and ML set by dividing logically a memory area into two are controlled by a processor P, and individual pairs of input data given in pair units are stored in memories MU and ML alternately while holding the input order. The processor P compares data stored in memories MU and ML in pair units and outputs data successively in accordance with sort conditions set by the algorithm of the merge sort method. If data compared with each other are equal to each other, the processor P gives such output control to a selector S that these data are outputted in the input order.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a sort processing device that can perform sort processing of large amounts of data quickly and efficiently.

[Technical background of the invention and its problems]

The sorting process that rearranges a collection of data into a predetermined order is a very important basic operation in the database processing process. However, since the data structure handled by a conventional general-purpose computer and the data structure handled by the database processing process are essentially different, it is recommended to use the above-mentioned general-purpose computer to perform the above-mentioned source processing on the large amount of data that makes up the database. This would require an extremely large amount of processing time. Therefore, in the past, various sorting processing apparatuses have been proposed which exclusively perform this type of sorting processing.

FIG. 1 is a schematic configuration diagram of a sorting processing device showing an example of the sorting processing device, in which a plurality of processing units (PEI, PE) are used to sort input data in predetermined units.

~PEn are connected in cascade. The PE at each stage consists of an internal memory in which two memory areas are logically set, and a processor that mutually compares the data stored in the two memory areas and controls the order in which they are read. be done. Then, under the control of C, the input data string is sequentially sorted in predetermined units, for example, according to the merge-sort algorithm.

By the way, conventionally 1. This type of sorting process is generally performed by sorting the data indicated by that key into outer layers, or descending order, according to one key. However, when dealing with a large-scale database, it is necessary to simultaneously perform sorting using a number of keys in multiple layers. For example, it is often desirable to classify a list by affiliation and to organize the names of people in the order of "iweo" for each affiliation. In such a case, conventionally, the list is generally sorted based on the affiliation key, and then the list is sorted using the person's name as a key for each of the classified affiliations. However, at this time, if the number of personnel belonging to a plurality of classified departments is different, it becomes necessary to provide a sorting program for each department according to the number of people. However, it is troublesome to detect the number of personnel belonging to each of the above-mentioned departments and set the number of repetitions of the sorting process accordingly. For this reason, simultaneous sort processing using multiple keys, so-called group pie processing, was extremely difficult) and was inefficient. Furthermore, if data is sorted according to the key and then the results are sorted using another key, there is a risk that the previous sorting result will be completely disturbed.

[Purpose of the invention]

The present invention has been made in consideration of these circumstances, and its purpose is to provide a practical method for easily, quickly, and efficiently sorting a large amount of data that constitutes a database according to multiple keys. Our goal is to provide a highly efficient sort processing device.

[Summary of the invention]

The present invention maintains the input order of input data and converts the input data to the first j? and a second memory, and these coarse data are compared using the Marzo sorting method, and the processors are sequentially output according to set sorting conditions, connected in cascade. , - The 'f processor of each stage is designed to read out the data to Q through the input order and output it, especially when the compared data are equal to each other. be.

Finally, if the compared data are equal and the sorting order by the key is not specified, the output order of the data is always the same as the input order.
This allows the results of sorting using other keys to be saved.

〔Effect of the invention〕

Thus, according to the present invention, if the compared data is large and the order of sorting by the key is not specified, the data is always output in the input order of the data, so the sorting process is performed first. The results of sorting performed according to other keys can be saved as is. Therefore, for example, in the case of sorting based on affiliation and name as in the example above, first sorting is done using name as the key, and then It becomes possible to easily and efficiently obtain the desired sorting results by simply executing the sorting process using the key . You can easily perform group pie processing. Therefore, instead of having to worry about setting the number of repetitions of sort processing according to the number of personnel classified by department as in the past, it is possible to effectively run a standardized processing program and easily and efficiently. This has great practical effects, such as being able to perform desired sorting processing.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

The apparatus of this embodiment is basically constructed by connecting a plurality of processing elements (PE2PEn) in cascade in the same manner as shown in FIG. In this embodiment, the processing stage PE is constructed as shown in FIG. 2. In FIG. 2, MU and ML are first and second memories set by logically dividing the memory area into two), and under the control of processor P, input data given in sets are Keep the input order and make sure to enter the first couplet in the couplet.
The memory MU1 and the second memory ML are alternately stored. The processor P processes the first and second memo! The data stored in JMU and ML are compared between the set units, and the compared data is output in order according to the set sorting conditions according to the algorithm of the merge sort method. Here, the output switching of the data is performed by the selector SK.

However, in the above merge sort method, for example, two sets of data 4
, 6.5.8 are stored in the first and second memories MU and ML respectively, first compare data 4 and 5 with each other, output data 4 from memory MU according to the condition of ascending order sorting, and then Compare data 6.5 to memory M
Output data 5 from L, further compare data 6.8 with each other, output data 6 from memory MU, output the last remaining data 8, and sort data strings 4, 5゜1, s. This is what you get as follows.

However, the input data set is 2.3°6.9.1
, 4, 6.9, according to the merge sort method described above, it is possible to obtain a data string of 1.2, 3, 4, 6°6.9, but among these, the data string of 6, 6 is obtained. It is not known from which memory MU or ML the data was read.

In other words, data 6 from either memo IJMU, ML
Even if the key is obtained, there is no problem with the sorting process related to that key. However, if the above data 6 is actually data 6&r6b with attributes based on other keys, some restrictions are often placed on the data output order, but in reality, in the sorting process based on the single key mentioned above, , the above constraints cannot be considered at all.

Therefore, in this device, from the processor P to the selector S
On the other hand, if the compared data are equal, the output order of the data is always the same as the input order. In other words, since the first set of data input is stored in the first memory MUK and the next set of data is stored in the second memory ML, the comparison data of these strings are equal to each other. In this case, the data stored in the first memo IJMU is always read out in order.

In the device configured in this manner, when the data "Q-s" consisting of data of multiple keys is repeatedly sorted with respect to the plurality of keys, the comparison data of the sorting processing by the keys is equal. When K.
Since the data will be output in the order in which it was input, the previously obtained sorting results will be saved as they are. Therefore, for example, two items consisting of affiliation and person's name.
If you want to perform group-by processing using one key,
If you sort the results using the name of the person as a key, which is a lower-level concept than shown in Figure 3, and then sort the results using the department as the key, you can save the results of sorting based on the person's name and then use the department as the key. It is possible to obtain the sorting results as follows. In other words, it is possible to sort the names of people in alphabetical order and then sort them into their affiliations, leaving the previous sorting results as they are for the names of each affiliation. For a specific example:),
Each piece of data is given as a data source consisting of (number, name, affiliation, address). These data are given in numerical order.

Therefore, we will now give a proposition to classify the above data by affiliation, and to arrange the names of each affiliation in alphabetical order. In this case, conventionally, attention has been paid to the affiliation of the input data indicated by &, and a sorting process is executed using this as a key to obtain a sorting result as indicated by bK in the figure. This makes it difficult to classify each piece of data by affiliation. However, since no sorting process has been performed on the names (names) K of the persons belonging to each organization, a sorting process using the names as keys is required next. However, since the number of personnel belonging to each department is different, it is necessary to prepare a sorting program for each department as described above, which is very troublesome.

In this regard, in this apparatus, the input data is sorted according to the concept of subclass, iJ) Sorting processing within classified affiliations, using the name as a key, according to ff+d#. That is, since the data is input in numerical order as shown in FIG.
As shown in Fig. 5, merge sort results are obtained in which the two data are grouped in sequence. For such a set of data, 2
The processing element PE in the third stage merge-sorts each set of data, and obtains the sorting result in which four data are grouped as shown in dK.Similarly, the processing element PK in the third stage ta merge-sorts these data sets. The sets of data are merge-sorted to obtain a sort processing result in which eight data sets are grouped as shown in eK. In this example, since the input data is 8 digits, the above 3-level masono ``-)K'' sorting process is completed.As a result, the input data is divided into 50 syllables J@ using the name as a key. It will be sorted.

After that, with respect to the results of the sorting process, f*g+
The busy sorting process is repeated as indicated by hK. However, in this case, in the first stage processing element PE1, the input data is sorted in units of 2, and 4 sets of merge sort results, each consisting of 2 sets of data, are obtained at 5, as shown by tK. After that, for such merge sort i-button, the second stage processing element p1. Now, a merge sort is performed in the same way, but at this time, two identical r-ri's with "design" as the key appear. In this case, in the past, it was fixed which data to output first, but with this device, the manually calculated data is always output in order. Therefore, the Maozon root processing result is ``Ishikawa Design'' and ``F'' as shown in f.
This means that the results of sorting by name, such as "Kamiya, Design", can be obtained in a state where they are saved as they are. Similarly, for the data ``Sakai\Design'' and ``Detection, Design,'' the results of sorting using the previous name in that group as a key are saved.゛And the third stage of processing Niremen)P
Even in Es, the key data is "sales" and "accounting"
When they are equal in [Design], the data is output in the order of input, so the results sorted in alphabetical order as shown in g are stored in each affiliation.

In this manner, sorting using two keys can be easily performed by simply repeating the sorting process using each key. Furthermore, unlike the conventional method, there is no need to rearrange the sorting process using the next key depending on the result of the sorting process performed first. By adding a simple control that inputs/outputs the data to [ if the comparison data is equal, the results of sorting by other keys can be effectively saved and sorting by multiple keys can be done. can be done effectively.

Furthermore, since the above processing can be repeatedly executed while performing the mother line processing, high-speed sorting processing is possible, and its practical advantage is enormous.

Note that the present invention is not limited to the above embodiments. For example, the number of stages of sort processing by the processing element PE may be determined according to the number of input data. Further, it is of course possible to carry out a sorting process using three or more keys, and the number of repetitions of the sorting process may be set according to the number of keys.

Furthermore, the configurations of the memories MU and ML and the processor P1 selector S may be determined according to specifications. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a basic configuration diagram of a sorting processing device, FIG. 2 is a configuration diagram of processing elements according to an embodiment of the present invention, FIG. 3 is a diagram showing the concept of a sorting process, and FIG. 4 is a diagram of a sorting process. It is a figure showing an example of an execution form. pE (pEl, pg!~Pchi)...processing element,
C... Controller, MU... First memory, ML
...Second memory, P...Zorosetsusa, S...
selector. Applicant: Director of the Agency of Industrial Science and Technology Kawa 1) Hirobe Figure 1 Figure 2 Figure 3 Figure 4

Claims (1)

[Claims] Input data given in units of sets is stored in first and second memories for each set while maintaining the input order, and the data of each set is compared with each other between the sets, and the results of this comparison and settings are the first or second sorting condition according to the sorting condition
It is configured by connecting a plurality of processing elements in cascade, each including a processor that sequentially reads data from the memories of the 2. A sorting processing device characterized in that the sorting processing device reads data in accordance with the order of input to the memory.