JP2011233012A - Merge processor, control program and record merge method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a merge processor, a control program and a record merge method for achieving the efficient input of records.SOLUTION: A merge processor 10 is provided with: an arrangement conversion part 20 for inputting a plurality of record columns consisting of a plurality of sorted records, and for dividing each record column into record blocks each of which includes at least one record, and for replacing a sequence so that a record for prediction included in each of the record blocks can be positioned at the front part; an input part 22 for inputting the record blocks to a storage part 24; and a record column selection part 26 for selecting the record column to be acquired based on the record for prediction. The input part 22 acquires the records in the order of the replacement of the sequence, and outputs the acquired record for prediction before acquiring all the records, and the record column selection part 26 selects the record column to be acquired next based on the record for prediction.

Description

  The present invention relates to a merge processing device, a control program, and a record merging method, and more particularly to a merge processing device using a prediction method, a control program for operating the merge processing device, and a record merging method using the merge processing device.

  In an information processing system such as a database, sorting is an important basic process, and a technique for processing the sorting at high speed is required. The merge sort method, which is a sort method, repeats in a stepwise manner a merge process that combines already obtained short sorted record strings to obtain a longer sorted record string. Thereby, it is possible to sort data obtained by collecting a large number of records.

  A configuration example of a commonly used merge processing apparatus will be described with reference to FIG. Terms used in the description are defined as follows with reference to Non-Patent Document 1. The rearrangement unit of the data to be sorted is “record”, and a plurality of records rearranged according to the sort order is “run”. The number of runs to be merged is expressed using an integer P of 2 or more. Each series is given a number (serial number) from 1 to P.

  The merge processing apparatus 100 inputs P series of record data to be merged from the external input supply unit 150, performs merge processing therein, and then obtains the merged series of records obtained from the external output storage unit 160. Output to. The merge processing apparatus 100 reads the input unit 110 to which a record is input from the input supply unit 150, the temporary storage unit 120 that stores the record input to the input unit 110, and the records in the temporary storage unit 120 one by one. A sort output unit 140 that rearranges them according to the sort order and outputs them to the external output storage unit 160. Further, an input ream selection unit 130 for selecting a ream input by the input unit is provided.

  Here, the temporary storage unit 120 does not have a capacity for storing all of the continuous records to be merged. Therefore, the temporary storage unit 120 proceeds with the merge process while acquiring the record data little by little. The record input process to the temporary storage unit 120 is performed with a record block that is a collection of one or more records as a unit. Further, the input unit 110 is shared by P series. For this reason, the temporary storage unit 120 cannot input records from P reams simultaneously, and advances the input while switching the reams to be input. As a specific example, there is a case where data of a record is read out from an SDRAM memory connected to the merge processing apparatus 100 via a memory bus using a burst transfer mode. There is Patent Document 1 as a document describing such cases. Alternatively, as another example, when record data is collectively input in units of blocks having a predetermined length from a disk device connected to the merge processing apparatus 100, a record of a certain length is stored from a storage server connected by Ethernet (registered trademark). There are cases where data is read in units of record blocks.

  Selection of the record block output from the input unit 110 to the temporary storage unit 120 during the merge process is performed by the input continuous selection unit. In order to advance the merge process at high speed without stopping the operation of the alignment output unit 140, the records are stored at a sufficient speed so that the number of records stored in the temporary storage unit 120 is not empty for all of the P series. It is necessary to input to the temporary storage unit 120. However, the speed at which records are output from the buffer used for each series varies depending on the input. Therefore, the input sequence selection unit 130 determines the input sequence while observing the record storage state of the temporary storage unit. As an effective technique in this input selection control, Non-Patent Document 1, P307 and Non-Patent Document 2, APPENDIX 2 have devised a selection method based on prediction.

  Here, the prediction disclosed in Non-Patent Document 1 and Non-Patent Document 2 serial numbers will be described in detail. For the i-th buffer in the temporary storage unit 120, the last record that has been input to the temporary storage unit 120 from the i-th sequence at that time is defined as the last record of the i-th sequence. In the prediction-based method, the input selection unit 130 selects a record that has the most forward sorting order from among the P last records, and selects the sequence that supplied the record as the sequence for inputting the next record. . The ream selected in this way indicates the ream that needs the earliest record supply at that time. Therefore, the input reaming selection method based on prediction is a preferable method for reducing the capacity of the record storage area required for the temporary storage unit 120 and the merge processing delay.

JP-A-5-298069

The Art of Computer Programming Volume 3 Sorting and Searching Second Edition Japanese Edition, Donald E. Knuth (Author), Makoto Arisawa (Director), Eiichi Wada (Director), ASCII, 2006 E.H.Friend, "Sorting on Electronic Computer Systems," J. ACM, vol. 3, no. 3, 1956, pp. 134-168.

  However, when using the input sequence selection based on prediction in the merge processing device, when the processing time required for the prediction itself or the delay time until the record data is actually input after determining the serial number of the input destination is large The record input process is not executed efficiently. Furthermore, there is a problem that the merge processing speed is lowered. This problem will be described with reference to FIG. FIG. 19 shows three execution order relationships: an input request process for the input supply unit 150, a process for inputting a record from the input supply unit 150, and a process for determining an input sequence. In this figure, it is assumed that there is a delay of 3 records from when the input supply unit 150 is requested to supply records to when record data is input. A record block as an input unit is composed of eight records, which are arranged in the order according to the sorting rule. When an input series is selected based on the prediction, the last input among these records is used. Therefore, as shown in FIG. 19, there is a problem that record input is not performed at all during the input series selection process and during input delay, and the efficiency of the record input process is poor.

  To solve this problem, there are methods such as preparing a redundant buffer and prefetching several record blocks, and speculatively selecting the next input series without waiting for all record inputs. It is done. However, the problem is that the required buffer capacity and processing delay time are increased, and a time penalty occurs when the prediction is lost.

  The present invention has been made to solve such a problem, and an object thereof is to provide a merge processing apparatus, a control program, and a record merging method capable of realizing efficient record input.

  In the merge processing device according to the first aspect of the present invention, a plurality of record strings composed of a plurality of sorted records are input, and each of the input record strings is converted into a record block including at least one record. An arrangement conversion unit that separates and changes the order of the records so that the prediction record arranged at the end of the plurality of records included in each of the record blocks is positioned ahead of the currently arranged position; The input unit obtains a plurality of record blocks from the arrangement conversion unit and inputs them to the storage unit, and the input unit is based on the size relationship between the prediction records included in each of the plurality of record blocks. A record string selection unit that selects a record string acquired from the layout conversion unit, and the input unit includes the record block. Each of the records included in the order of the records in which the order has been changed, before acquiring all the records included in the record block, output the acquired prediction record to the record string selection unit, The record sequence selection unit starts processing for selecting a record sequence that the input unit acquires from the layout conversion unit next in response to the input of the prediction record output from the input unit.

  In the merge processing apparatus according to the second aspect of the present invention, a plurality of record strings each including a plurality of sorted records are input, and each of the input record strings includes a plurality of records including at least one record. The record block is obtained in order from the separation unit that is separated into blocks, and the record block that is input later to the separation unit, and the record block is input to the storage unit, and the block located at the forefront in the record block is used for prediction An input unit that is extracted as a record, and a record string selection unit that selects a record block that the input unit acquires from the arrangement conversion unit, based on the magnitude relationship of the prediction records included in each of the plurality of record blocks; The records are sequentially retrieved from the storage unit according to the sort order opposite to the sort order of the record sequence. An alignment output unit that outputs and outputs the prediction record to the record sequence selection unit before acquiring all the records of the record block, and the record sequence selection unit Based on the prediction record output from the input unit, the input unit next starts a process of selecting a record string acquired from the arrangement conversion unit.

  In the merge processing device according to the third aspect of the present invention, a plurality of record strings composed of a plurality of sorted records are input, and each of the input record strings is converted into a record block including at least one record. An arrangement in which the order of the records is changed so that the first prediction record arranged at the end of the plurality of records included in each of the record blocks is positioned ahead of the currently arranged position. A conversion unit, a first input unit that obtains a plurality of record blocks from the arrangement conversion unit and inputs them to the first storage unit, and from the first storage unit according to the same sort order as the sort order of the record sequence A first alignment output unit for sequentially extracting and outputting records, and the record block sequentially input from the record block to the layout conversion unit; A second input unit that obtains a first block and inputs the second block to the second storage unit, and extracts a block arranged at the foremost position in the record block as a second prediction record; and a sort order of the record sequence The second sort output unit that sequentially extracts and outputs records from the second storage unit according to the reverse sort order and the magnitude relationship between the first prediction records included in each of the plurality of record blocks Based on the first record sequence selection unit that selects the record sequence that the first input unit acquires from the arrangement conversion unit, and the second prediction record included in each of the plurality of record blocks A second record string selection unit that selects a record string that the second input unit acquires from the arrangement conversion unit based on a magnitude relationship; The second input unit outputs the first and second prediction records to the first and second record sequence selection units before acquiring all the records of the record block, and The second record string selection unit is configured such that, based on the first and second prediction records output from the first and second input units, then the first and second input units are the arrangement conversion unit. The process of selecting a record string to be acquired from is started.

  A control program according to a fourth aspect of the present invention includes: a storage unit that stores a plurality of record strings including a plurality of records; and a record that is extracted from the records stored in the storage unit and rearranged in an order according to a sorting rule. Based on the prediction output arranged at the end at the time of input from the outside in each of the record blocks, the aligned output unit that outputs and a plurality of record sequences are separated into record blocks including at least one record, A control program for causing a computer to execute a process of inputting the record string in a merge processing unit including a record string selection unit that selects a record string to be input next, and acquiring a plurality of the record blocks from the outside And the schedule arranged at the end of the plurality of obtained record blocks. Changing the order of the records so that the records for use are positioned ahead of the currently arranged position, and inputting the records to the merge processing unit in the order of the records whose order has been changed. To be executed.

  In the record merging method according to the fifth aspect of the present invention, a plurality of record strings composed of a plurality of sorted records are input, and each of the input record strings is converted into a record block including at least one record. The order of the records is changed so that the prediction record arranged at the end of the plurality of records included in each of the plurality of record blocks is positioned ahead of the currently arranged position. Acquiring each of the records included in the record block in the order of the records in which the order has been changed, outputting the acquired prediction record, and depending on the input of the output prediction record And a step for starting a process for selecting a record string to be input next. Is shall.

  According to the present invention, it is possible to provide a merge processing apparatus, a control program, and a record merging method capable of realizing efficient record input.

1 is a configuration diagram of a merge processing apparatus according to a first embodiment; 1 is a configuration diagram of a merge processing apparatus according to a first embodiment; It is a figure which shows the rearrangement of the record concerning Embodiment 1. FIG. 3 is a flowchart of prediction processing according to the first exemplary embodiment. It is a figure which shows the operation | movement of the record input request concerning Embodiment 1. FIG. FIG. 3 is a configuration diagram of a merge processing apparatus according to a second embodiment; It is a figure which shows the rearrangement of the record concerning Embodiment 2. FIG. FIG. 10 is a configuration diagram of a merge processing apparatus according to a third embodiment; It is a figure which shows the rearrangement of the record concerning Embodiment 3. FIG. It is a figure which shows the rearrangement of the record concerning Embodiment 3. FIG. It is a figure which shows the rearrangement of the record concerning Embodiment 4. FIG. It is a figure which shows the rearrangement of the record concerning Embodiment 4. FIG. FIG. 10 is a configuration diagram of a merge processing apparatus according to a fifth embodiment; FIG. 10 is a configuration diagram of a merge processing apparatus according to a sixth embodiment; FIG. 10 is a configuration diagram of a merge processing apparatus according to a sixth embodiment; FIG. 10 is a configuration diagram of a sort processing apparatus according to a seventh embodiment; FIG. 10 is a configuration diagram of a sort processing apparatus according to an eighth embodiment; It is a block diagram of a general merge processing apparatus. It is a figure which shows the operation | movement of the record input request in a common merge processing apparatus.

(Embodiment 1)
Embodiments of the present invention will be described below with reference to the drawings. A configuration example of the merge processing apparatus according to the first exemplary embodiment of the present invention will be described with reference to FIG. The merge processing apparatus 10 includes an arrangement conversion unit 20, an input unit 22, a storage unit 24, and a record string selection unit 26.

  The arrangement conversion unit 20 receives a plurality of record strings composed of a plurality of sorted records. Furthermore, the arrangement conversion unit 20 separates the input plurality of record strings into record blocks including at least one record, and the prediction record arranged at the end of the plurality of records included in each record block The order of the records is changed so as to be positioned in front of the currently arranged position.

  The input unit 22 acquires a plurality of record blocks from the arrangement conversion unit 20 and inputs them to the storage unit 24. The input unit 22 acquires from the layout conversion unit 20 the records included in the record block in the order of the records whose order has been changed. Thereby, the input part 22 can acquire the record for prediction, before acquiring all the records contained in a record block. The input unit 22 outputs the acquired prediction record to the record string selection unit 26 before acquiring all the records included in the record block.

  The record string selection unit 26 selects a record string that the input unit 22 acquires from the arrangement conversion unit 20 based on the prediction records included in each of the plurality of record blocks. The record sequence selected by the record sequence selection unit 26 is acquired from the layout conversion unit 20 by the input unit 22 after the acquisition of the current record sequence in the input unit 22 is completed.

  The storage unit 24 stores the record string input from the input unit 22.

  As described above, by using the merge processing device 10 described in FIG. 1, the input unit 22 can acquire a prediction record before acquiring all the records included in the record block. Thereby, the input unit 22 can output the prediction record to the record sequence selection unit 26 before acquiring all the records included in the record block. Therefore, before the input unit 22 acquires all the records included in the record block from the arrangement conversion unit 20 based on the output prediction record, the record sequence selection unit 26 next inputs the arrangement conversion unit. The process of selecting the record string acquired from 20 can be started. Therefore, since the time from when the record sequence is input to the input unit from the layout conversion unit until the next record sequence is input can be shortened, efficient input processing of the record sequence can be realized. .

  Subsequently, a detailed configuration example of the merge processing apparatus according to the first exemplary embodiment of the present invention will be described with reference to FIG. The merge processing apparatus 10 includes an arrangement conversion unit 20, an input unit 22, a storage unit 24, a record string selection unit 26, an input storage unit 30, and an alignment output unit 32. The storage unit 24 includes a first buffer 41, a second buffer 42, and a Pth buffer 43. The merge processing apparatus 10 is connected to the input supply unit 50 and the output storage unit 60. Components of the merge processing apparatus 10 having the same functions as those in FIG. 1 will be described with the same reference numerals as those in FIG. In addition, the term “run” used in the following description has the same meaning as the record string described above.

  The merge processing apparatus 10 receives P series of record data to be merged from an external input supply unit 50. Each series is identified using a number from 1 to P (serial number). The merge processing apparatus 10 performs merge processing of P series in the apparatus, and then outputs the obtained merged series of records to the external output storage unit 60.

  The arrangement conversion unit 20 acquires the records supplied from the input supply unit 50, and rearranges the arrangement of the records. For example, the arrangement conversion unit 20 divides a plurality of records included in the series into record blocks having a predetermined number of records, and sets the record arranged at the end of the record block, that is, the prediction record, to the top. Arrange to position.

  The input storage unit 30 receives records from the layout conversion unit 20 and stores the sorted records in units of record blocks.

  The input unit 22 acquires records from the input storage unit 30 in the order of the records whose order has been changed by the arrangement conversion unit 20. The input unit 22 acquires a record for each record block from the sequence selected by the record string selection unit 26. When the input unit 22 acquires the prediction record, the input unit 22 outputs the acquired prediction record to the record sequence selection unit 26 before acquiring all the records included in the record block. Further, the input unit 22 restores the acquired records to the sort order before rearrangement and outputs the records to the storage unit 24.

  The storage unit 24 has a plurality of buffers for storing records for each series. For example, when P runs are input to the merge processing apparatus 10, the buffer that stores run 1 is the first buffer 41, the buffer that stores run 2 is the second buffer 42, and the run P is stored. The buffer is a Pth buffer 43. The storage unit 24 stores the records in the respective buffers in the order of the restored records for each series.

  The alignment output unit 32 reads the records in the storage unit 24 according to the sort order, and outputs the records to the external output storage unit 60. For example, the alignment output unit 32 reads the record having the smallest record value located at the head of each buffer. Thereby, the alignment output unit 32 can sort the records in order from the record with the smallest value and merge a plurality of sequences.

  The record string selection unit 26 selects a record block acquired by the input unit 22 from the input storage unit 30 by a method based on prediction. Below, the method based on prediction is demonstrated.

  When the input unit 22 inputs a record block to the storage unit 24 to the record sequence selection unit 26, a prediction record included in each record block is output from the input unit 22, and the record sequence selection unit 26 Holds a total of P prediction records output last, one for each series. The record string selection unit 26 extracts a record in which the sort order is the front, that is, a record having the smallest value, from among the P prediction records. The record string selection unit 26 selects a series supplied with the extracted record as a series for inputting the next record. The ream selected in this way indicates the ream that needs the earliest record supply at that time. The record string selection unit 26 notifies the selected ream to at least one of the input unit 22 and the input storage unit 30.

  When the input selected by the record string selection unit 26 is notified, the input storage unit 30 outputs the record blocks included in the selected sequence to the input unit 22. Or the input part 22 may request | require the output of the record block contained in the selected continuous with respect to the input memory | storage part 30, if the series selected by the record sequence selection part 26 is notified.

  When the input unit 22 acquires the record block from the input storage unit 30, the input unit 22 outputs the prediction record to the record sequence selection unit 26, and the record sequence selection unit 26 and the other continuous records that it holds. The record with the smallest value is extracted using the prediction record.

  As described above, the record block acquired by the input unit 22 from the input storage unit 30 is selected by a method based on prediction.

  Next, the timing of the processing operation in the merge processing apparatus 10 will be described. Here, the input unit 22, the storage unit 24, the alignment output unit 32, and the input continuous selection unit 26 in the merge processing apparatus 10 are collectively referred to as a merge processing unit. When the merge process is started, necessary initialization is performed in each unit. Thereafter, the arrangement conversion unit 20 waits for record data input from the external input supply unit 50. The layout conversion unit 20 starts to rearrange the record data and start storing the record data in the input storage unit when the input of the record data is started. The arrangement conversion unit 20 continues processing until input of all record data is completed. In the middle of the process, when a sufficient amount of record data is accumulated in the input storage unit 30, the merge processing unit starts the merge process. The merge processing unit continues the process until all records are output to the external output storage unit 60. The operation timing of the arrangement conversion unit 20 and the merge processing unit may operate synchronously while maintaining a certain time difference, or may operate at independent timing. However, if the record data to be processed has not been input, the merge processing unit waits without performing processing. Alternatively, when the input storage unit 30 has a capacity capable of storing all the record data, the merge processing unit starts the processing after completing the accumulation of the record data input from the arrangement conversion unit 20 You may do it.

  Next, a record rearrangement process according to the first embodiment of the present invention will be described with reference to FIG. 3, a1, a2, a3,... Represent arbitrary continuous records supplied from the outside, and the subscript numbers represent the order according to the sorting rule. Each series of records is supplied to the arrangement conversion unit 20 in accordance with the sort order, as shown in FIG. The arrangement changing unit 20 divides these records into record blocks having eight records such as a1 to a8, a9 to a16, and a17 to a24. Further, the arrangement converting unit 20 positions the records in the record block by placing the last record (prediction record) of each record block at the beginning of the record block as shown in FIG. Rearrange as you do. The arrangement conversion unit 20 stores the series whose record arrangement has been changed in this way in the input storage unit 30 in units of record blocks.

  FIG. 3B shows that the last record (a8, a16, a24...) Is arranged at the top in each record block, and the other records are stored with one position shifted backward. It is shown. The record data is supplied from the input storage unit 30 to the input unit 22 in units of record blocks, and the records in the record block are supplied in this rearranged order.

  The input unit 22 inputs the records to the storage unit 24 while returning the record arrangements in the record blocks rearranged by the arrangement conversion unit 20 to the original arrangement order. As a result, even if records are input in an arrangement different from the sorted order, the records are stored in the storage unit 24 in the sorted order. Therefore, the alignment output unit 32 can execute the merge process by comparing and extracting the top records of the respective buffers.

  Next, the flow of merge processing according to the first embodiment of the present invention will be described with reference to FIG. First, when merge processing is started in the merge processing unit, the record string selection unit 26 initializes variables to be held (S1). For example, the record string selection unit 26 initializes the serial number of the prediction record having the minimum value and the minimum value of the prediction record.

  Next, the storage unit 24 determines whether there is a vacancy in the buffer of the sequence i (S2). When it is determined that there is a vacancy in the buffer of the sequence i, the record string selection unit 26 determines whether the prediction record of the sequence i is the smallest among the input prediction records (S3). When the record string selection unit 26 determines that the prediction record for the sequence i is the minimum, the record sequence selection unit 26 updates the variable indicating the sequence number i and the minimum value of the prediction record (S4). Next, the processing of steps S2 to S4 is similarly repeated for the sequence i + 1 (S5).

  If the storage unit 24 determines in step S2 that there is no space in the buffer of the sequence i, or if the record sequence selection unit 26 determines in step S3 that the record for prediction of the sequence i is not the minimum, The process proceeds to S5. Next, the record string selection unit 26 determines whether or not the processes after step S2 have been executed for all the series (S6). When it is determined by the record string selection unit 26 that all the reams have not executed the processes after step S2, the process proceeds to step S2. When it is determined by the record string selection unit 26 that all the sequences have executed the processing from step S2 onward, it is determined whether or not there is an empty buffer in the storage unit 24 (S8). When it is determined that there is an empty buffer, the record string selection unit 26 notifies the updated serial number to at least one of the input unit 22 and the input storage unit 30 and makes a record input request (S9). ). If the record string selection unit 26 determines that there is no free buffer, the process ends.

  As described above, by using the merge processing apparatus according to the first embodiment of the present invention, record block input processing from the input storage unit 30 to the input unit 22 can be continuously executed. The effect in Embodiment 1 of this invention is demonstrated using FIG. The input unit 22 acquires records whose order has been rearranged by the arrangement conversion unit 20. For example, when the prediction conversion record is arranged by the arrangement conversion unit 20 so as to be positioned at the head of the record block, the input unit 22 selects the prediction record as the record string selection unit before acquiring all the records of the record block. 26 can be output. Thereby, before the input part 22 acquires all the records, the record sequence selection part 26 selects the series to input next, and makes a record input request to at least one of the input part 22 and the input storage part 30. Thereby, there is a time margin for concealing the delay time from when the record input request is made until the input unit 22 actually acquires the record from the input storage unit 30. Therefore, the input storage unit 30 can continuously execute record input processing on the input unit 22.

(Embodiment 2)
Next, a configuration example of the merge processing apparatus according to the second exemplary embodiment of the present invention will be described with reference to FIG. The merge processing device 10 in FIG. 6 differs from the merge processing device in FIG. 2 in that the arrangement conversion unit 20 is not provided. The other configurations in FIG. 6 are the same as those in FIG. 2 and will be described using the same reference numerals as those in FIG.

  The input storage unit 30 has a capacity sufficient to store all the merged records from the input supply unit 50. Therefore, when the input storage unit 30 receives all record data from the input supply unit 50, the merge processing by the merge processing unit is started. The input storage unit 30 stores records while maintaining the sort order of records input from the input supply unit 50. Here, a1, a2,..., And the order of input from the records with the smallest values are assumed to be ascending. The reverse order is the descending order. The input storage unit 30 separates the record data input from the input supply unit 50 into record blocks including at least one record. A record block having records a1 to a8 is referred to as a record block 1, a record block having records a9 to a16 is referred to as a record block 2, and a record block having records a41 to a48 is referred to as a record block 6. Record blocks 3 to 5 are set in a17 to a41. FIG. 7A shows an arrangement example in the case where the number of records in the record block is 8 and a series having 48 records is stored in the input storage unit 30.

  Unlike the first embodiment of the present invention, the input unit 22 acquires record blocks from the input storage unit 30 in the order of record block 6, record block 5,... Record block 1 (descending order). However, the input unit 22 acquires the records in the record block in ascending order. Further, the input unit 22 stores the acquired records in the storage unit 24 while rearranging all the records in the record block in descending order. FIG. 7B shows the order of records acquired by the input unit 22 from the input storage unit 30 and the order of records stored in the storage unit 24.

  Here, the prediction record used by the record sequence selection unit 26 for prediction is the last record of the record block when the sort order is descending, that is, the forefront record of the record block when the sort order is ascending To do. The input unit 22 acquires the records in the record block in ascending order. Therefore, the prediction record is first acquired by the input unit 22 among the records in the record block. When the input unit 22 acquires the prediction record, the input unit 22 outputs the acquired prediction record to the record string selection unit 26 before all the records in the record block are input. Thereby, before the input part 22 acquires all the records in a record block, the record sequence selection part 26 starts the process which determines the series of the record block acquired next. In Embodiment 2 of the present invention, the record string selection unit 26 selects a series of prediction records for which the maximum value is set from among the prediction records.

  As described above, in the merge processing device according to the second exemplary embodiment of the present invention, the input unit 22 moves from the input storage unit 30 at the same timing as the merge processing device according to the first exemplary embodiment of the present invention. Records can be obtained. Therefore, it is possible to increase the moving efficiency of the merge processing unit and increase the processing speed of the merge processing.

(Embodiment 3)
Next, a configuration example of the merge processing apparatus according to the third embodiment of the present invention will be described with reference to FIG. The merge processing apparatus according to the third embodiment of the present invention has a configuration combining the first and second embodiments of the present invention. The input unit 22, the storage unit 24, the record string selection unit 26, and the alignment output unit 32 are assumed to be an ascending order merge processing unit, and the ascending order merge processing unit is the same as the merge processing unit according to the first embodiment of the present invention. To work. Further, the input unit 22_1, the storage unit 24_1, the record string selection unit 26_1, and the alignment output unit 32_1 are set as a descending order merge processing unit, and the descending order merge processing unit is a merge processing unit according to the second embodiment of the present invention. The same operation is performed.

  Similar to the merge processing apparatus according to the second embodiment of the present invention, the input storage unit 30 has a capacity enough to store all the merged record data. Therefore, merge processing by the ascending order merging processing unit and descending order merging processing unit is started after all record data is input from the arrangement conversion unit 20 to the input storage unit 30. The record input from the input supply unit 50 is input to the arrangement conversion unit 20 as in the merge processing apparatus 10 according to the first embodiment of the present invention. The layout conversion unit 20 starts with the records input from the input supply unit 50 in the sorted order as shown in FIG. 9A and the last record of each record block as shown in FIG. 9B. Change the order of the records so that The arrangement conversion unit 20 inputs a plurality of rearranged records to the input storage unit 30.

  The input storage unit 30 notifies the control unit in the merge processing apparatus of the total number of input records or the storage capacity occupied by the record data, and the control unit writes the merged record sequence in the output storage unit 60 based on this. Reserve space. When the reservation is completed, a command is issued to start the merge process to the ascending order merge processing unit and the descending order merge processing unit.

  The input unit 22 acquires records in the order of record block 1, record block 2,. The subsequent processing is the same as that of the merge processing apparatus according to the first embodiment of the present invention as shown in FIG. 10A, and the input unit 22 restores the records in the order before they are rearranged. To the storage unit 24. These records are rearranged in ascending order and then sequentially output to the output storage unit 60. The output record is written from the beginning of the area secured in the output storage unit 60.

  The input unit 22_1 acquires records in the order of the record block 6, the record block 5,. Further, in the descending order merge processing unit, the prediction records are a1, a9,..., A41 as in the second embodiment of the present invention. In Embodiment 3 of the present invention, since the order of records is changed in the arrangement conversion unit 20, the prediction record is arranged second in each record block. When acquiring the prediction record, the input unit 22_1 outputs the prediction record to the record string selection unit 26_1 before acquiring all the records in the record block. In the subsequent processing, as in the merge processing apparatus according to the second embodiment of the present invention, the input unit 22 sorts the records in descending order and inputs them to the storage unit 24_1. These records are rearranged in descending order and then sequentially output to the output storage unit 60. The output records are written in reverse order from the rear of the area secured in the output storage unit.

  Here, the ascending order merge processing unit performs the merge processing in order from the record block located in front of the input series. On the other hand, the descending order merging processing unit performs the merging processing in order from the record block located behind the input series. Thus, the input continuous records may be controlled so that the merge processing is executed by either the ascending order merge processing unit or the descending order merge processing unit. For example, the number of records output by the ascending order merge processing unit and the number of records output by the descending order merge unit are monitored by the control unit in the merge processing apparatus, and the sum of both numbers is added to the total number of records input. It may be performed by stopping the operation of both units when it reaches. Thereby, it is possible to prevent duplicate records from being written to the output storage unit 60.

  As described above, by using the merge processing apparatus according to the third embodiment of the present invention, merge processing by the ascending order merge processing unit and merge processing by the descending order merge processing unit are performed on the input P series. Can be executed in parallel. Therefore, compared with Embodiments 1 and 2 of the present invention, high-speed merge processing can be realized.

(Embodiment 4)
Subsequently, a record replacement pattern of the arrangement conversion unit 20 according to the fourth embodiment of the present invention will be described. The configuration of the merge processing apparatus according to the fourth embodiment of the present invention is the same as the configuration of FIG. The record replacement pattern of the arrangement conversion unit 20 according to the fourth embodiment of the present invention will be described with reference to FIG.

  FIG. 11A shows the order of the records of the series 1 input to the arrangement conversion unit 20. On the other hand, FIG. 11B shows the order of the records in the series 1 after the records are replaced by the arrangement conversion unit 20. The record block 1 will be described as an example. The order of records in the record block before being rearranged by the arrangement conversion unit 20 is arranged in the order of a1 to a8. The order of the records in the record block 1 after being rearranged by the arrangement conversion unit 20 is rearranged in the order of a8, a1, a7, a2, a6, a3, a5, a4. Records are rearranged as follows. First, let a1 to a4 in the record block 1 before being rearranged be the first half, and let a5 to a8 be the second half. Here, the records in the latter half are rearranged in descending order. Next, the first half of the records arranged in ascending order and the second half of the records arranged in descending order are alternately arranged. As a result, the above-described record order is obtained.

  FIG. 12A shows the rearrangement of records performed when records are input from the input unit 22 to the storage unit 24. As in FIG. 10A, the records are rearranged in ascending order. Is input to the storage unit 24. FIG. 12B shows record rearrangement performed when records are input from the input unit 22_1 to the storage unit 24_1. As in FIG. 10B, the records are arranged in descending order. It is changed and inputted to the storage unit 24.

  As described above, as shown in FIGS. 12A and 12B, when the first half and the second half of the record block are alternately arranged, the input unit 22_1 is more than the case described in FIG. 10B. The delay time from when the record is input to the storage unit 24_1 can be shortened. In FIG. 10B, the input unit 22_1 cannot output a record to the storage unit 24_1 until the record a47 that reaches the eighth is acquired. On the other hand, in FIG. 12B, the input unit 22_1 can output the record to the storage unit 24_1 when acquiring the fifth record a46. Accordingly, the output delay time to the storage unit 24_1 in the descending order merge processing unit can be shortened. Although the output delay time to the storage unit 24_1 in the ascending order merge processing unit increases, it is possible to prevent the delay time in one merge processing unit from significantly increasing.

(Embodiment 5)
Subsequently, a configuration example of the merge processing apparatus according to the fifth exemplary embodiment of the present invention will be described with reference to FIG. The merge processing apparatus shown in FIG. 13 is a pipeline merge sort apparatus in which N merge processing units are connected in series. The merge processing apparatus is an apparatus for merging a P number of records column, the first merge processing unit 71, a second merge processing unit 72, ..., merge processing unit 73 of the N, respectively P ₁ , P ₂ ,..., _PN are provided with a function of merging record strings. Here, it is assumed that P = P ₁ × P ₂ ×... × _PN , and that each merge processing unit includes a sufficient input storage unit for storing a record string. The first merge processing unit 71 starts merge processing when P ₁ sorted record sequences are input from the outside, and outputs the merged record sequence to the input storage unit of the second merge processing unit 72, This is repeated P ÷ P ₁ = P ₂ ×... × P _N times. The second merge processing unit 72, a record sequence from the first merge processing unit 71 is input _two P starts merging process, the records columns merged into the input storage unit of the third merge processing unit Output, and repeat this for P ₃ ×... × P _N times. In this way, this merge processing apparatus merges P record strings.

  In this merge processing apparatus, each merge processing unit has the same configuration as that of the merge processing apparatus according to the second embodiment of the present invention. When N = 1, it is the same as the second embodiment of the present invention. If the sort order of the record sequence input to the first merge processing unit is ascending order, the sort sequence of the record sequence output from the first merge processing unit 71 to the second merge processing unit 72 is descending order. Similarly, the sort order of the record sequence output from the second merge processing unit 72 to the third merge processing unit is ascending order, and the Nth merge processing unit 73 inputs the ascending order record sequence when N is an odd number. A descending record string is output, and when N is an even number, a descending record string is input and an ascending record string is output.

  The internal operation of the merge processing unit is as described in the second embodiment. The input storage unit included in each processing unit needs to read in record block units, and further causes a read delay time. However, by switching the input sort order and the output sort order so that the prediction record can be read at the beginning of the record block according to the output sort order, the read delay is concealed and the record input efficiency is improved. And enables high-speed operation.

(Embodiment 6)
Next, a configuration example of the merge processing apparatus according to the sixth embodiment of the present invention will be described with reference to FIG. The merge processing apparatus shown in FIG. 14 is also a pipeline merge sort apparatus having the same configuration as that of the fifth embodiment, but includes arrangement conversion units 74 and 75 between each merge processing unit, and N pieces The internal configuration and operation of the merge processing unit are different. Here, the first merge processing unit 71 is a general merge processing device, and the second to N-th merge processing units are combined with the arrangement conversion unit in front of the first merge processing unit 71 according to the first or third embodiment of the present invention. The same configuration and operation as in FIG. For example, FIG. 15 shows a configuration when N = 2 and the second merge processing unit 72 has the same configuration as that of the first embodiment of the present invention. In this merge processing apparatus, the sort order of the input record string and the sort order of the output record string are the same in the arrangement conversion unit and all the merge processing units.

  Here, the arrangement conversion unit 74 between the first merge processing unit 71 and the second merge processing unit 72 performs arrangement conversion while separating records sequentially output from the first merge processing unit 71 into record blocks. At the same time, the converted record block data is written to the second merge processing unit 72. Since this arrangement conversion process is performed simultaneously with the output of the merge processing unit, the increase in the merge processing time due to the arrangement conversion process is small. Similarly, the arrangement conversion unit between the Nth merge processing unit 73 and the (n + 1) th merge processing unit operates simultaneously with the record output of the Nth merge processing unit 73, and therefore performs the arrangement conversion process. There is little increase in merge processing time. Therefore, the merge processing apparatus according to the embodiment of the present invention increases the efficiency of record input and enables high-speed operation.

(Embodiment 7)
Next, a configuration example of the sort processing apparatus according to the seventh embodiment of the present invention will be described with reference to FIG. This sort processing apparatus is a sort apparatus in which a sort processing unit 81 is added to the input of the merge processing apparatus according to the fifth embodiment of the present invention. The sort processing unit 81 inputs a plurality of records to an internal storage unit, divides them into groups of P records, and performs sorting within the group. The sorted record string is output to the first merge processing unit 71. The sort order in the sort processing unit 81 is the reverse of the sort order in the first merge processing unit. The sort processing apparatus configured as described above operates at a high speed similarly to the merge processing apparatus of the fifth embodiment.

(Embodiment 8)
Next, a configuration example of the sort processing apparatus according to the eighth embodiment of the present invention will be described with reference to FIG. This sort processing apparatus is a sort apparatus in which the first merge processing unit 71 in the merge processing apparatus according to the sixth embodiment of the present invention is replaced with a sort processing unit. The sort processing unit 81 inputs a plurality of records to an internal storage unit, divides them into groups of P records, and performs sorting within the group. The sorted record string is output to the first merge processing unit 71. The sort processing apparatus configured as described above operates at a high speed similarly to the merge processing apparatus according to the sixth embodiment.

  In the first to eighth embodiments described above, the number of records included in the input ream (record sequence) is described as being equal for ease of explanation, but the number of records is not limited. . In other words, the number of records included in the input series is arbitrarily determined for each series.

  An application example of the merge processing apparatus according to the first to eighth embodiments of the present invention will be described below.

  First, an example in which the merge processing apparatus according to the first to eighth embodiments is configured by an integrated circuit will be described. The input storage unit 30 and the alignment output unit 32 are composed of SDRAM and its control circuit, and the storage unit 24 is composed of SRAM and its control circuit. The SDRAM can perform high-speed data reading by using the burst transfer mode, and the number of records in the record block is set large so that the reading efficiency from the SDRAM is high. The address where the record block is arranged is aligned with an appropriate multiple of addresses. There is an access delay from the issuance of a command for reading data to the SDRAM until the data is transferred, which causes a reduction in the efficiency of record input processing in the merge processing apparatus. However, since the merge processing device 10 of the present invention rearranges the arrangement when storing the record blocks in the SDRAM, the next read request processing is completed before all the records in the record block are transferred to the input unit 20. Can be performed on SDRAM. Therefore, the access delay until data is transferred from the SDRAM can be hidden. Thus, when the merge processing device is configured by an integrated circuit, the merge processing can be executed faster than the merge device not using the present invention.

  Next, an example in which the merge processing device according to the first to eighth embodiments is configured by a computer and a hard disk device including a CPU and a main memory, and a program for causing the computer to perform merge processing will be described. A hard disk is used for storing records in the input storage unit 30 and the alignment output unit 32, and a main storage memory inside the computer is used for the storage unit 24. This computer can operate a plurality of threads in parallel, and this program activates a plurality of threads in the arrangement changing unit 20, the input unit 22, the record string selection unit 26, and the alignment output unit 32, and the CPU Describes a command for performing the merge processing according to the present invention. When the record data stored in the hard disk device is read, there is a delay time from when the read command is issued until the data is read. This causes a reduction in the efficiency of the record input process in the merge processing device. However, the merge processing apparatus of the present invention rearranges the arrangement when storing the record blocks on the hard disk. Therefore, the next read request process can be performed on the hard disk before all the records in the record block have been transferred to the input unit 22. Therefore, the access delay until data is read from the hard disk device can be hidden. Thereby, when the merge processing device is configured by a computer device, the merge processing can be executed at a higher speed than the merge processing device not using the present invention.

In the above-described embodiments, the present invention has been described as a hardware configuration, but the present invention is not limited to this. The present invention can also realize arbitrary processing by causing a CPU (Central Processing Unit) to execute a computer program.

  The program may be stored using various types of non-transitory computer readable media and supplied to a computer. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (for example, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (for example, magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / W and semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (random access memory)) are included. The program may also be supplied to the computer by various types of transitory computer readable media. Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

  Note that the present invention is not limited to the above-described embodiment, and can be changed as appropriate without departing from the spirit of the present invention.

  (Supplementary note 1) A plurality of record strings composed of a plurality of sorted records are input, and each of the input record strings is separated into record blocks including at least one record and included in each of the record blocks. A record conversion unit that changes the order of the records so that the prediction record arranged at the end of the plurality of records is positioned ahead of the currently arranged record, and a plurality of record blocks from the arrangement conversion unit And selecting a record string that the input unit acquires from the layout conversion unit based on the size relationship between the input unit that inputs to the storage unit and the prediction record included in each of the plurality of record blocks A record string selection unit that performs the recording, and the input unit precedes each of the records included in the record block. Before acquiring all the records included in the record block, in order of the records whose order has been changed, the acquired prediction record is output to the record sequence selection unit, and the record sequence selection unit The merge processing apparatus which starts the process which the said input part next selects the record row | line | column acquired from the said arrangement | positioning conversion part according to the input of the record for prediction output from the input part.

  (Supplementary note 2) The merge processing device according to supplementary note 1, wherein the arrangement conversion unit changes the order of the records so that the prediction record is positioned at a head of the record block.

  (Supplementary note 3) In the supplementary note 1 or 2, the input unit restores the order of the record blocks in which the order of the records has been changed, and inputs the record block in which the order has been restored to the storage unit. The merge processing device described.

  (Supplementary Note 4) A separation unit that inputs a plurality of record strings composed of a plurality of sorted records and separates the input record strings into a plurality of record blocks including at least one record, and the separation The record block is sequentially input from the record block that is input later to the unit, and the record block is input to the storage unit, and the input unit that takes out the block positioned at the forefront in the record block as a prediction record; Based on the magnitude relationship of the prediction records included in each of the record blocks, the input unit selects a record block to be acquired from the arrangement conversion unit, and the record sequence is reverse to the sort sequence of the record sequence A sort output unit that sequentially extracts records from the storage unit according to the sort order and outputs them The input unit outputs the prediction record to the record sequence selection unit before acquiring all the records of the record block, and the record sequence selection unit is based on the prediction record output from the input unit. Then, a merge processing device that starts processing for selecting a record string that the input unit acquires from the layout conversion unit.

  (Supplementary note 5) A plurality of record strings composed of a plurality of sorted records are input, and each of the input record strings is separated into record blocks including at least one record and included in each of the record blocks A first conversion record arranged at the end of the plurality of records to be arranged, the arrangement conversion unit for changing the order of the records so as to be positioned in front of the currently arranged position, and a plurality of the arrangement conversion unit from the arrangement conversion unit A first input unit that inputs to the first storage unit, and sequentially outputs records from the first storage unit according to the same sort order as the record sequence. The record block is acquired in order from the record block input afterward to the alignment output unit and the arrangement conversion unit, and the second storage And the second input unit that takes out the block arranged at the foremost position in the record block as a second prediction record, and the second input unit according to the sort order opposite to the sort order of the record sequence. Based on the magnitude relationship between a second alignment output unit that sequentially extracts and outputs records from the storage unit, and the first prediction record included in each of the plurality of record blocks, the first input unit includes: Based on the magnitude relationship between a first record sequence selection unit that selects a record sequence to be acquired from the layout conversion unit and the second prediction record included in each of the plurality of record blocks, the second A second record sequence selection unit that selects a record sequence acquired from the arrangement conversion unit, and the first and second input units are configured to record the record block. Before acquiring all the records of the first and second records, the first and second record records are output to the first and second record string selectors, and the first and second record string selectors are A process of selecting a record sequence that the first and second input units acquire from the arrangement conversion unit next based on the first and second prediction records output from the first and second input units Merge processing device to start.

  (Supplementary note 6) The merge processing device according to supplementary note 5, wherein the arrangement conversion unit alternately arranges a first half column of records included in the record block and a column of records in which the remaining records are rearranged in reverse order.

  (Supplementary note 7) A storage unit that stores a plurality of record strings composed of a plurality of records, an alignment output unit that takes out the records stored in the storage unit and rearranges them in an order according to a sorting rule, and a plurality of output units The record sequence is separated into record blocks including at least one record, and the record sequence to be input next is selected based on the record for prediction arranged at the end of each record block at the time of external input. A control program for causing a computer to execute a process of inputting the record string in a merge processing unit including a record string selection unit, the step of acquiring a plurality of the record blocks from the outside, and the plurality of acquired records The record for prediction placed at the end of the block is currently placed A step of replacing the order of the records to be located forward of the location, a control program for executing the steps, to a computer for inputting the record in the order of records in which the sequence has been replaced with the merge processing unit.

  (Supplementary Note 8) A storage unit that stores a plurality of record strings composed of a plurality of records, an output unit that outputs records by taking out the records stored in the storage unit and rearranging them in an order according to a sorting rule, The record sequence is separated into record blocks including at least one record, and the record sequence to be input next is selected based on the record for prediction arranged at the end of each record block at the time of external input. A control program for causing a merge processing unit including a record sequence selection unit to execute a process of inputting the record sequence on a computer, the step of acquiring a plurality of the record blocks from the outside, and the plurality of acquired record blocks Sorting the records contained in and the sorted plurality Changing the order of the records so that the prediction record arranged at the end of the record block is positioned ahead of the currently arranged position; and the records in the order of the records in which the order is changed A control program for causing a computer to execute the step of inputting to the merge processing unit.

  (Additional remark 9) The step which isolate | separates each said input record row | line | column into the record block containing at least 1 record from which the several record row | line | column comprised from the some sort | sorted record is input, and the said several record block Reordering the records so that the prediction record arranged at the end of the plurality of records included in each of the records is positioned ahead of the position where the records are currently arranged, and the records included in the record block Are obtained in the order of the records in which the order has been changed, and a step of outputting the obtained prediction record and a record string to be input next are selected in accordance with the input of the output prediction record And a record merging method comprising:

  (Supplementary Note 10) A plurality of record sequences each composed of a plurality of records are input, and each of the input record sequences is separated into record blocks including at least one record; and records included in the record block are The order of the records is changed so that the prediction record arranged at the end of the plurality of records included in each of the plurality of record blocks is positioned ahead of the currently arranged position. Acquiring each of the records included in the record block in the order of the records in which the order has been changed, outputting the acquired prediction record, and depending on the input of the output prediction record , Start the process to select the next input record string Record merge method comprising Tsu and up, the.

  (Additional remark 11) A plurality of record sequences composed of a plurality of records are input, and each of the input record sequences is separated into a plurality of record blocks including at least one record, and records included in the record block are The sort processing unit for sorting, and the record blocks are acquired in order from the record block input later to the sort processing unit, input to the storage unit, and the block located at the forefront in the record block is used for prediction An input unit that is extracted as a record, and a record string selection unit that selects a record block that the input unit acquires from the arrangement conversion unit, based on the magnitude relationship of the prediction records included in each of the plurality of record blocks; The storage unit according to the sort order opposite to the sort order of the record sequence. An alignment output unit that sequentially extracts and outputs the records, and the input unit outputs the prediction records to the record sequence selection unit before acquiring all the records of the record block, and the record sequence selection unit Is a sort processing device that starts processing for selecting a record sequence that the input unit acquires from the arrangement conversion unit next, based on the prediction record output from the input unit.

  (Supplementary note 12) A plurality of record strings composed of a plurality of records are input, the input record strings are separated into record blocks including at least one record, and the records included in the record block are sorted. Arrangement conversion that rearranges the order of records so that the sorting record and the prediction record arranged at the end of the plurality of records included in each of the record blocks are positioned ahead of the currently arranged position A plurality of record blocks from the layout conversion unit and input to the storage unit, and the input unit based on the magnitude relationship of the prediction records included in each of the plurality of record blocks A record string selection unit that selects a record string to be acquired from the arrangement conversion unit, and the input unit , Each of the records included in the record block is acquired in the order of the records in which the order is changed, and before acquiring all the records included in the record block, the acquired prediction records are stored in the record string selection unit. The record sequence selection unit sorts to start processing for selecting a record sequence that the input unit acquires from the layout conversion unit in response to the input of the prediction record output from the input unit Processing equipment.

DESCRIPTION OF SYMBOLS 10 Merge processing apparatus 20 Arrangement conversion part 22, 22_1 Input part 24, 24_1 Storage part 26, 26_1 Record sequence selection part 30 Input storage part 32, 32_1 Alignment output part 41 1st buffer 42 2nd buffer 43 3rd P buffer 50 Input supply unit 60 Output storage unit

Claims (10)

A plurality of record sequences composed of a plurality of sorted records are input, and each of the input record sequences is separated into record blocks including at least one record, and a plurality of records included in each of the record blocks An arrangement conversion unit that changes the order of records so that the prediction record arranged at the end of the record is positioned ahead of the currently arranged position;
An input unit that obtains a plurality of record blocks from the arrangement conversion unit and inputs to the storage unit;
Based on the magnitude relationship of the prediction records included in each of the plurality of record blocks, a record sequence selection unit that selects a record sequence that the input unit acquires from the arrangement conversion unit, and
The input unit is
Each of the records included in the record block is acquired in the order of the records whose order has been changed, and before the acquisition of all the records included in the record block, the acquired prediction records are transferred to the record sequence selection unit. Output,
The record string selection unit
The merge processing apparatus which starts the process which the said input part next selects the record row | line | column which the said input part acquires from the said arrangement | positioning conversion part according to the input of the record for prediction output from the said input part. The arrangement conversion unit
The merge processing apparatus according to claim 1, wherein the order of the records is changed so that the prediction record is positioned at a head of the record block. The input unit is
The merge processing apparatus according to claim 1, wherein the order of the record blocks in which the order of the records is changed is returned to the original, and the record block in which the order is returned is input to the storage unit. A plurality of record strings composed of a plurality of sorted records, and a separation unit that separates each of the input record strings into a plurality of record blocks including at least one record;
The record block is sequentially input from the record block input later to the separation unit, and input to the storage unit, and the input unit that takes out the block located most forward in the record block as a record for prediction,
Based on the magnitude relationship of the prediction records included in each of the plurality of record blocks, a record string selection unit that selects a record block that the input unit acquires from the arrangement conversion unit;
An alignment output unit that sequentially extracts and outputs records from the storage unit in accordance with a sort order opposite to the record sequence sort order;
The input unit is
Before acquiring all the records of the record block, output the prediction record to the record string selection unit,
The record string selection unit
A merge processing apparatus that starts processing for selecting a record sequence that the input unit acquires from the arrangement conversion unit next based on the prediction record output from the input unit. A plurality of record sequences composed of a plurality of sorted records are input, and each of the input record sequences is separated into record blocks including at least one record, and a plurality of records included in each of the record blocks An arrangement conversion unit that changes the order of the records so that the first prediction record arranged at the end of the first prediction record is positioned ahead of the position at which the current prediction is arranged;
A first input unit that obtains a plurality of record blocks from the arrangement conversion unit and inputs them to a first storage unit;
A first alignment output unit that sequentially extracts and outputs records from the first storage unit according to the same sort order as the record sequence;
The record blocks are acquired in order from the record block input later to the arrangement conversion unit, input to the second storage unit, and the block most forwardly arranged in the record block is second predicted. A second input part to be taken out as a record for use,
A second alignment output unit for sequentially extracting and outputting records from the second storage unit in accordance with a sort order opposite to the record sequence sort order;
A first record string selection that selects a record string that the first input unit acquires from the arrangement conversion unit based on the magnitude relationship of the first prediction records included in each of the plurality of record blocks And
Second record sequence selection for selecting a record sequence that the second input unit acquires from the layout conversion unit based on the magnitude relationship of the second prediction records included in each of the plurality of record blocks And comprising
The first and second input units are:
Before acquiring all the records of the record block, the first and second prediction records are output to the first and second record string selection units,
The first and second record string selection units are:
Next, based on the first and second prediction records output from the first and second input units, the first and second input units select a record sequence to be acquired from the arrangement conversion unit. Merge processing device that starts processing. The arrangement conversion unit
The merge processing apparatus according to claim 5, wherein a first half column of records included in the record block and a column of records in which the remaining records are rearranged in reverse order are alternately arranged. A storage unit that stores a plurality of record strings composed of a plurality of records, an alignment output unit that outputs records by taking out the records stored in the storage unit and rearranging them in an order according to a sorting rule, and a plurality of record sequences A record string selection unit that selects a record string to be input next based on a prediction record that is separated into record blocks including at least one record and is arranged at the end of each record block at the time of external input. A control program for causing a computer to execute a process of inputting the record string in a merge processing unit including:
Obtaining a plurality of the record blocks from the outside;
Reordering the records so that the prediction records arranged at the end of the plurality of obtained record blocks are positioned ahead of the currently arranged positions;
A control program for causing a computer to execute the step of inputting the records to the merge processing unit in the order of the records in which the order has been changed. A storage unit that stores a plurality of record strings composed of a plurality of records, an alignment output unit that outputs records by taking out the records stored in the storage unit and rearranging them in an order according to a sorting rule, and a plurality of record sequences A record string selection unit that selects a record string to be input next based on a prediction record that is separated into record blocks including at least one record and is arranged at the end of each record block at the time of external input. A control program for causing a computer to execute a process of inputting the record string in a merge processing unit including:
Obtaining a plurality of the record blocks from the outside;
Sorting the records included in the plurality of obtained record blocks;
Reordering the records so that the prediction records arranged at the end of the plurality of sorted record blocks are positioned ahead of the currently arranged positions;
A control program for causing a computer to execute the step of inputting the records to the merge processing unit in the order of the records in which the order has been changed. A step of inputting a plurality of record strings composed of a plurality of sorted records, and separating each of the input record strings into record blocks including at least one record;
Reversing the order of the records so that the prediction record arranged at the end of the plurality of records included in each of the plurality of record blocks is positioned ahead of the currently arranged position;
Obtaining each of the records included in the record block in the order of the records in which the order is changed, and outputting the acquired prediction record;
Starting a process of selecting a record string to be input next in response to the input of the output prediction record. A plurality of record strings composed of a plurality of records are input, and each of the input record strings is separated into record blocks including at least one record;
Sorting the records contained in the record block;
Reversing the order of the records so that the prediction record arranged at the end of the plurality of records included in each of the plurality of record blocks is positioned ahead of the currently arranged position;
Obtaining each of the records included in the record block in the order of the records in which the order is changed, and outputting the acquired prediction record;
Starting a process of selecting a record string to be input next in response to the input of the output prediction record.

Images