JP3555506B2 - Character string data compression encoding device, character string data decompression device, and character string data arithmetic processing device - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a character string data compression encoding apparatus, a character string data decompression apparatus, and a character string data arithmetic processing apparatus. In particular, the present invention relates to data communication and processing, and compression encoding of character string data for efficient transmission or storage. The present invention relates to a character string data compression encoding apparatus, a character string data decompression apparatus, and a character string data arithmetic processing apparatus for performing high-speed arithmetic processing between compression-encoded character string data.
[0002]
[Prior art]
A Lempel-Ziv encoding method is widely known as a data compression method for encoding an input character string as a copy of a character string that appeared in the past. Also, a number of patents have been filed relating to the improved system. For example, Japanese Patent No. 2590287 describes an invention relating to a system in which the Lempel-Ziv coding system is improved.
[0003]
In general, the Lempel-Ziv encoding method has a memory for storing (part of) the history of a character string input in the past, and a dictionary for registering / retrieving a subsequence included in the memory. Are sequentially read, and a partial sequence is registered in a dictionary, and the same partial sequence is detected with reference to this to perform data compression. There are two Lempel-Ziv encoding schemes: a universal scheme and an incremental decomposition scheme.
[0004]
In the universal Lempel-Ziv encoding method, when an input character string is encoded, a character string portion having the maximum length that matches the current character string is separated from an arbitrary position in the past input character string (character string substring). This is a method of performing data compression by designating the position of the character substring in the memory and the length of the character substring.
The incremental classification type Lempel-Ziv encoding method encodes and registers a character string obtained by adding a newly appearing character to a previously encoded maximum-length character string, and stores the current character string in the past. This is a method of encoding as a copy of the longest character subsequence of the subsequence.
[0005]
In any of the methods, inventions and improvements have been made with the object of increasing the compression ratio of character string data and efficiently performing compression / decompression processing. That is, it is assumed that the code string data that has been compression-encoded is once returned to the original character string data through the restoration processing, and then subjected to various arithmetic processing. Performing direct arithmetic processing on compression-encoded code string data without going through decompression processing is not considered in the prior art, and is a code devised so that such arithmetic processing can be performed efficiently. Not.
[0006]
On the other hand, as a method of compressing data, storing the data in a storage device, and performing high-speed arithmetic processing in a state in which the data is compressed, an expression method and an arithmetic method of logical function data using a data structure called a “Binary Decision Graph” are known. Are known. The processing method is described in R. E. FIG. Bryant: "Graph-Based Algorithms for Boolean Function Manipulation", IEEE Trans. compute. , Vol. C-35, no. 8, pp. 677-691 (hereinafter referred to as article 1). In addition, a detailed commentary article is also published in “Special Issue BDD (Binary Decision Graph)” (hereinafter referred to as “Article 2”), pp. 584-630, May 1993, IPSJ Journal “Information Processing”.
[0007]
The logical function data is a logical function data indicating whether a logical value of 0 or 1 is output as a function value when a combination of logical values of 0 and 1 is given to a plurality of input variables. It represents information.
Before the invention of the binary decision diagram, to represent logic function data,ⁿA method of expressing a function value for a combination of input variable values (n is the number of input variables) in a form of a sequence of 0s and 1s called a “truth value table” is often used. FIG.₁, X₂, X₃3 is an example of a truth table expressing logic function data having the three variables as inputs.
[0008]
The BDD is a method of expressing logic function data using a graph,
Means for giving a fixed order to all input variables (hereinafter referred to as ordering processing);
Means for deriving two partial functions from a logical function by substituting logical values of 0 and 1 into one of the input variables (hereinafter referred to as expansion processing);
The above expansion processing is repeated according to the order given by the ordering processing until there is no input variable to which 0 and 1 are substituted, and an intermediate node representing each expansion processing and two partial functions obtained by each expansion processing are obtained. A binary tree graph as shown in FIG. 13 is generated using 0-branch and 1-branch points, and 0-terminal nodes and 1-terminal nodes representing logical values of 0 and 1 obtained as a result of repeating the expansion processing. Means (hereinafter referred to as binary tree generation processing),
In the above-described binary tree generation processing, as shown in FIG. 14, it is checked whether or not 0-branches and 1-branches of two intermediate nodes in the binary tree point to the same node. Means for finding two intermediate nodes that are equivalent to each other and erasing one of the intermediate nodes and sharing the other intermediate node (hereinafter referred to as sharing processing);
At each intermediate node, when the 0-branch and the 1-branch both point to the same node as shown in FIG. 15, means for removing the intermediate node and directly connecting the branch (hereinafter referred to as non-redundancy processing) Call),
Consists of
[0009]
By performing the above processes in combination, a binary decision diagram as shown in FIG. 16 is obtained. This BDD can be stored in the storage device in the form of a table listing intermediate nodes as shown in FIG. The amount of storage required at this time is proportional to the number of nodes in the BDD. When the logical function data is represented by a truth table, the number of combinations of the values of the input variables (when the number of variables is n, 2ⁿIs always required, but using a binary decision diagram often results in a smaller storage amount, and in some cases, the compression ratio can reach several hundred times. For this reason, binary decision diagrams have been widely used in recent years as a compression storage method for logical function data.
[0010]
Further, the paper 1 and the article 2 store a binomial operation processing method between the BDDs. That is, when two BDDs are stored in the storage device, a binary logical operation (logical function, OR, exclusive OR, etc.) between the logic function data represented by those BDDs is executed. Then, a method is described in which a new BDD expressing the logical function data of the operation result is generated and stored in a storage device.
[0011]
The operation method is such that by assigning 0 and 1 to the uppermost variables of two given BDDs, each is decomposed into two corresponding sets of intermediate nodes, and the obtained set of intermediate nodes is further divided. Decompose and repeat this to get multiple sets of terminal nodes. A new terminal node representing the binomial operation result for each of the terminal node sets is newly generated, and the operation result obtained in this manner is assembled again into the binary decision diagram, thereby generating a new binary decision diagram. At this time, by registering the set of intermediate nodes that appear during the arithmetic processing and the operation result for the set in the dictionary, when the same set of intermediate nodes is detected, the set is not further decomposed and immediately registered in the dictionary. There is provided a high-speed means for referring to an intermediate node as a calculation result.
[0012]
The calculation time of this calculation method is almost proportional to the sum of the number of nodes of the two BDDs given as the calculation objects and the number of nodes of the BDD generated as the calculation result. Therefore, as compared with the logical operation method using the truth table data format, the effect of speeding up the arithmetic processing can be obtained in substantially proportion to the data compression ratio by the BDD.
[0013]
[Problems to be solved by the invention]
In the Lempel-Ziv encoding method, it is assumed that the code string data that has been compression-encoded is once returned to the original character string data through the restoration processing, and then subjected to various arithmetic processing. By performing the compression encoding, the amount of transmission data or the storage area can be saved, but processing time for compression / decompression is additionally required. It is generally considered that the compression ratio and the processing time have a trade-off relationship.
[0014]
However, if the arithmetic processing can be directly performed without decompressing the compression-encoded code string data, processing time for compression and decompression becomes unnecessary, and the arithmetic processing is performed in proportion to the data compression ratio. It would be possible to shorten the time.
For this purpose, it is necessary to devise a method of dividing the sub-sequence in the compression encoding so that the arithmetic processing is easy. However, in the conventional Lempel-Ziv encoding method, substrings are separated at different positions depending on the contents of input character string data. Therefore, when trying to perform a binomial operation process between two compressed code strings, However, there is a problem in that the correspondence relationship between the subsequence data becomes complicated, and it is not possible to efficiently execute the arithmetic processing with the data compressed.
[0015]
On the other hand, in the binomial logical operation processing method using the binary decision diagram, the binomial arithmetic processing can be executed in the data format as it is, and the speeding up which is almost proportional to the compression ratio is realized.
However, in the binary logical operation processing method using the binary decision diagram, the two binary decision diagrams to be operated are delivered to the arithmetic processing device in a state stored as a node table in the storage device, and the binary decision graph of the operation result is obtained. Is also generated as a node table in the storage device. Therefore, a requirement is that all nodes of these BDDs can be stored in the storage device in advance. In the process of sharing a BDD and the process of non-redundancy described in the section of the related art, a method of registering and retrieving all nodes in a dictionary in a storage device is used in order to quickly determine whether a node matches. Have been. The capacity of the dictionary is finite, and if the BDD of the operation result does not fit in the dictionary, the process ends abnormally in the middle of the process and there is a problem that the result of the operation process cannot be obtained.
[0016]
Further, in the binary operation processing method using the BDD, the calculation process cannot be started until all nodes of the two BDDs to be operated are stored in the storage device. That is, the processing of the next stage cannot be started unless the processing of the previous stage is completed. Therefore, there is a problem that continuous arithmetic processing cannot be connected in series and executed in parallel in a pipeline manner.
[0017]
SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and has been developed in view of compression-coded code string dataNot necessarily completelyA character data compression / encoding device, a character string data restoring device, and a character string data processing device capable of directly performing arithmetic processing without decompression and enabling high-speed arithmetic processing according to a data compression ratio To provideWith the goal.
Also, the present invention does not start the operation after storing all the input data in the storage device as in the conventional operation processing method using the BDD, but instead starts the input compressed code string data. It is an object of the present invention to provide a character string data compression / encoding device, a character string data decompression device, and a character string data arithmetic processing device capable of sequentially executing arithmetic processing while sequentially reading from the same, and sequentially outputting the operation result.
[0018]
[Means for Solving the Problems]
FIG. 1 is a diagram showing the principle configuration (compression encoding device) of the present invention.
The present invention (claim 1)ⁿ(Where n is an integer constant equal to or greater than 0) A character string data compression encoding apparatus which receives as input character string data and outputs code string data obtained by compressing the data without losing the information. ,
Input processing means 110 for sequentially reading the character string data before compression from the beginning,
Output processing means 120 for sequentially outputting the code string data after compression encoding from the beginning,
A finite capacity dictionary 130;
Length input from input processing means 1102 ⁿ Is divided into two substrings each having a half length of the first half and the second half of the input character string data, and divided into the first half and the second half of the obtained substring. 2,4,8,16, ... 2ⁿA decomposition rule 150 that is referenced to decompose into
While sequentially reading the input character string data input from the input processing means 110 with reference to the decomposition rule 150, the lengths 1, 2, 4, 8, 6,.ⁿIs obtained by assigning a different registration number to each different character string, storing it in a dictionary as long as its capacity permits, and reading the input character string data while referring to the dictionary to obtain the input character string. Detecting means 141 for detecting that the same character string appears repeatedly in the column data a plurality of times;
If a substring consisting of the same character string appears repeatedly in the input character string data a plurality of times, the output processing unit 120 is instructed to output a set of the character string and the registration number of the character string at the first time. And an output control unit 142 for instructing the output processing unit 120 to output only the registration number indicating the subsequence after the second time.
[0019]
According to the present invention (claim 2), when character string data of an arbitrary length is input, the length of the input character string data is exactly two.ⁿThere is provided a means for adding a plurality of special characters representing blanks to the end of the character string data so that (where n is an integer constant as small as possible).
According to the present invention (claim 3), in the detection means 141,
An arbitrary character string a having a length of 2 or more is decomposed into two partial strings a1 and a2 each having a half length of the first half and the second half, and the character strings of the partial strings a1 and a2 are If it is already stored in the dictionary, the character string a is represented by a set of registration numbers indicating the character strings of the subsequence a1 and the subsequence a2, and a new registration representing the character string a in the set Storage means for giving a number and storing the number in the dictionary 130;
In the output control means 142,
When a character string a having a length of 2 or more is output, if the character strings of the two partial strings a1 and a2 of the first half and the second half are not stored in the dictionary 130, the partial strings a1 and The character string a is expressed by sequentially outputting the character strings in the column a2, and if one or both of the partial sequence a1 and the partial sequence a2 are already stored in the dictionary 130, the stored character string a is stored. Means is provided for outputting a registration number indicated by the character string instead of outputting the character string.
[0020]
According to the present invention (claim 4), in the storage means,
When the capacity of the dictionary 130 reaches the limit and a new character string x cannot be stored, it is still used as a first half or a second half of another character string in the character strings stored so far. Search for a character string y that has not been deleted, delete a set of registration numbers indicating the first half and the second half of the character string y representing the character string y, and point to the same registration number as the deleted character string y. There is means for storing information representing the new character string x in the indicated storage area, and advancing the subsequent processing.
[0021]
FIG. 2 is a block diagram showing the principle of the present invention (character string data restoration device).
The present invention (claim 5) is a character string data restoring device for restoring and extracting original character string data from code string data compressed and encoded by the character string data compression encoding device,
An input processing means 210 for sequentially reading the compression-encoded code string data from the beginning,
Output processing means 220 for sequentially outputting the restored character string data from the beginning,
A dictionary 230 having a capacity equal to or greater than the dictionary of the character string data compression encoding device used when compressing the character string data,
When sequentially reading the compression-encoded code string data obtained from the input processing means 210, if a partial string expressed as a simple character string is read, the output processing means is configured to output the character string as it is. When the instruction is given to the subroutine 220 and the set of the registration number and the character string of the subsequence is read, the character string is stored in the dictionary 230 with the registration number added thereto. Has a restoration control unit 240 that instructs the output processing unit 220 to extract the subsequence indicated by the registration number from the dictionary 230 and output it.
[0022]
According to the present invention (claim 6), in the restoration control means 240,
When sequentially reading the compression-encoded code string data, when a character string represented by a set of two partial strings of the first half and the second half is input, the first half and the second half of the partial string are input. Means for expressing the character string with a set of registration numbers and storing it in the dictionary 230;
When extracting and outputting the partial sequence stored in the dictionary 230, the partial sequence of the first half and the second half of the partial sequence is sequentially referred to from the dictionary 230, and the partial sequence is further divided into the first half and the second half. In the case of being expressed as a set, there is further provided a means for sequentially referring to them and extracting and outputting a subsequence of length 1.
[0023]
FIG. 3 is a diagram showing the principle configuration (character string data processing device) of the present invention.
According to the present invention (claim 7), when a certain binary operation for calculating and outputting one character for any two characters is defined, two character strings having the same length are used. What is claimed is: 1. A character string data processing device that outputs data of one character string representing a result of calculating a predetermined binary operation with respect to characters at the same position from the beginning as input data,
First input means for sequentially reading the compression-encoded code string data from the beginning, first output means for sequentially outputting the restored character string data from the beginning, and character string data used when the character string data is compressed When sequentially reading the first dictionary having a capacity equal to or larger than the dictionary of the compression encoding device and the compression-encoded code string data acquired from the first input means, a short simple character string is used. When the expressed substring is read, the first output means is instructed to output the character string as it is, and when the set of the registration number and the character string of the substring is read, the When the character string is given the registration number in the first dictionary and stored, and only the registration number of the subsequence is read, the subsequence indicated by the registration number is taken out from the first dictionary and output. To the first output means That restoration control becomes unit and two strings data restoration device having, two input processing means 310 and 320 for transferring to restore each input code string data to the character string data,
Arithmetic processing means 330 for sequentially calculating a binary operation between characters at the same position from the beginning of two character string data transferred from input processing means 310 and 320 and transferring the result as new character string data When,
A second input unit for sequentially reading the character string data before compression from the beginning, a second output unit for sequentially outputting the code string data after compression and encoding from the beginning, a second dictionary having a certain finite capacity, Length 2 input from the second input meansⁿIs divided into two substrings each having a half length of the first half and the second half of the input character string data, and divided into the first half and the second half of the obtained substring. 2,4,8,16, ... 2 ⁿ Decompose into subsequences ofReferenced toWhile sequentially reading the decomposition rule and the input character string data input from the second input means with reference to the decomposition rule, the lengths 1, 2, 4, 8, 16,. ⁿ , A different registration number is assigned to each different character string, and stored in the second dictionary as long as its capacity permits. The input character string data is stored while referring to the second dictionary. By reading, a detecting means for detecting that the same character string repeatedly appears in the input character string data a plurality of times, and a subsequence consisting of the same character string in the input character string data being repeated a plurality of times If it appears, it instructs the second output means to output a set of a character string and a registration number of the character string the first time, and outputs only the registration number indicating the subsequence after the second time Output control means for instructing the second output means to perform the output, wherein the character string data transferred from the arithmetic processing means 330 is compression-encoded and output as code string data. Processing means 36 With the door.
[0024]
In the present invention (claim 8), in the input processing means 310, 320 and the output processing means 360, when a subsequence is stored or referred to in each of the first dictionary or the second dictionary, the registration number of each registration number is determined. Further comprising a transfer means for transferring information;
The information of the registration number is acquired from the transfer means, and which subsequence corresponds to the result of the binomial operation with respect to a certain set of two input subsequences is determined by three input subsequences and one output subsequence. As a set of the registration numbers of the subsequences, as long as the capacity permits, the data is stored in a certain finite-capacity operation acceleration dictionary, and the operation processing is sequentially performed with reference to the operation acceleration dictionary. Detecting means for detecting that the set has appeared more than once;
When the same set of input subsequences appears, the input processing means 310 and 320 control so as to omit the process of restoring the registration number of the subsequence to character string data, and the output processing means 360 calculates And control means for controlling so as to output the registration number of the subsequence of the operation result stored in.
[0025]
The compression encoding device of the present inventionⁿ(Where n is an integer constant of 0 or more) character string data is input, and code string data obtained by compressing the data is output without losing the information. According to the compression encoding device, the length 2ⁿIs divided into two substrings each having a half length of the first half and the second half of the input character string data, and the obtained substring is further divided into the first half and the second half respectively. , 1,2,4,8,16, ..., 2^(N-1)Is decomposed into a number of subsequences.
[0026]
Further, based on the above decomposition rules, while sequentially reading the input character string data, the lengths 1, 2, 4, 8, 16,.^(N-1)Is extracted, a different registration number is assigned to each different character string, and the dictionary is stored as long as its capacity is allowed. By reading the input character string data while referring to this dictionary, it is possible to detect that the same character string appears repeatedly in the character string data a plurality of times. If a substring consisting of the same character string appears repeatedly in the input character string data a plurality of times, the first time a set of the character string and its registration number is output, and the second and subsequent registrations indicate the character string. By outputting only the number, the data is compression-coded.
[0027]
Further, in the compression encoding apparatus of the present invention, the restriction on the length of the input character string data is removed, that is, the length of the input character string data is exactly 2ⁿ(However, n is an integer constant as small as possible), a plurality of special characters representing blanks are added to the end of the character string data, and these are compression-encoded.
Further, in the compression encoding apparatus of the present invention, the lengths 1, 2, 4, 8, 16,.^(N-1)Is not enumerated and stored in the dictionary as it is, but is decomposed into two subsequences of the first half and the second half, and stored. That is, when the character string a having a length of 2 or more is decomposed into two partial strings a1 and a2 each having a half length of the first half and the second half, and the character strings a1 and a2 are already stored in the dictionary. Represents a character string a with a set of registration numbers indicating two character strings, and a new registration number representing the character string a is given to the character string a and stored in the dictionary. As a result, all the subsequences having a length of 2 or more are represented by a set of registration numbers indicating the subsequences of the first half and the second half.
[0028]
When a character string a having a length of 2 or more is output, if the character strings of the two partial strings a1 and a2 of the first half and the second half are not stored in the dictionary, the two character strings are sequentially output. To represent the character string a, otherwise (ie, if one or both of the strings a1 and a2 are already stored in the dictionary), instead of outputting the stored string, By outputting the registered registration number, the data can be compressed and encoded.
[0029]
In the compression encoding apparatus of the present invention, in the process of decomposing the character string data into subsequences and storing them in the dictionary, if the capacity of the dictionary reaches the limit and a new character string x cannot be stored, A character string y that has never been used as a first half or a second half of another character string is searched for from among the character strings stored up to and including information representing the character string y (the first half of the character string). And a registration number indicating the latter half of the character string), temporarily store information representing the new character string x in the storage area indicated by the same registration number as the deleted character string y, and proceed with the subsequent processing. .
[0030]
The character string data restoration device of the present invention, when sequentially reading code string data compressed and encoded by the data compression encoding device, when reading a partial string represented by a simple character string, the character string is left as it is. Output, when the set of the registration number and the character string of the subsequence is read, the registration number is added to the character string and stored in the dictionary, and when only the registration number is read, the part indicated by the registration number The character string data is restored by extracting the string from the dictionary and outputting it.
[0031]
Further, the character string data restoring device of the present invention decomposes the character string data into two partial strings, a first half part and a second half part, and restores the compression-encoded data. That is, when sequentially reading the compression-encoded code string data, if a character string represented by a set of two partial strings of the first half and the second half is input, the partial string of the first half and the second half is input. Having a set of registration numbers, the character string is represented and stored in the dictionary.On the other hand, when extracting and outputting the subsequence registered in the dictionary, the first half and the second half of the subsequence are sequentially read from the dictionary. If these subsequences are further expressed as a pair of the first half and the second half, the sequence is repeatedly referred to, and a subsequence of length 1 is finally extracted and output. To output the entire subsequence.
[0032]
The character string data arithmetic processing device of the present invention sequentially reads two code string data compressed by the data compression coding apparatus and restores these compressed code strings to obtain two characters that would be obtained. With respect to the column data, one character string data representing the result of performing a predetermined binary operation is calculated for characters at the same position from the beginning, and these are sequentially output in a state of being compressed by the compression method. . Further, the code string data input by the two input processing means are respectively restored to character string data and transferred to the arithmetic processing means, and the two character strings transferred from the input processing means are transferred to the arithmetic processing means. The binary operation of the characters at the same position from the beginning of the data is sequentially calculated, and the result is transferred to the output processing unit as new character string data. The output processing unit transfers the result from the arithmetic processing unit. The compressed character string data is compression-encoded and output as code string data.
[0033]
Further, the character string data arithmetic processing device of the present invention is provided with a detection means, a control means, and an arithmetic high-speed dictionary having a certain finite capacity, and the two input processing means and the output processing means each have a partial sequence. When stored in the dictionary or referred to, the information of each registration number is transferred to the detecting means. In the detection means, the capacity of the operation speedup dictionary is determined as a set of registration numbers of the three subsequences, as a set of registration numbers of the three subsequences. As long as the same set of input subsequences appears a plurality of times, the calculation is performed sequentially while referring to this dictionary. When the same set of input subsequences appears in the control means, the process of restoring the registration numbers of the subsequences to character string data in the input processing means is omitted, and the output processing means stores them in the high-speed operation dictionary. By outputting the registration number of the substring of the operation result as it is, it is possible to speed up the operation process.
[0034]
As described above, in the present invention, the character string data is divided into two parts, the first half part and the second half part, instead of cutting out the partial string at different positions depending on the contents of the input character string data as in the Lempel-Ziv encoding method. , And each subsequence is further divided into two subsequences, a first half and a second half, and this is repeated to obtain a length of 1, 2, 4, 8, 16,.^(N-1)Is used. Then, while sequentially reading the input data from the head, the subsequences are taken out, and while the data similar to the BDD are sequentially read from the head, the subsequences are taken out, and the data structure similar to the BDD is obtained. And register it in the dictionary.
[0035]
Adopting such a fixed decomposition method may reduce the chance of detecting a matching sub-sequence and reduce the compression ratio as compared with a case where a free break position is allowed. However, by fixing the delimiter position as described above, regardless of the content of the character string data, two times from the beginning of the data^kSince the position corresponding to a multiple of (1 <k <n) is always the boundary of the subsequence, the correspondence between the respective subsequences of the plurality of compressed code sequences is simplified.
[0036]
Therefore, in a process of inputting two compressed code strings as input and outputting a binary operation result of the character strings as a new compressed code string, a result of the binary operation for a certain set of two input subsequences is output. To store which sub-sequence of the code sequence corresponds to a set of registration numbers of these three sub-sequences in the high-speed operation dictionary as far as the capacity permits, and to sequentially perform the operation processing with reference to this dictionary Thus, it is possible to detect that the same set of input subsequences appears multiple times. At the second and subsequent appearances, by outputting only the dictionary registration number indicating the subsequence of the operation result, it is possible to save the trouble of repeating the same operation process and to increase the speed.
[0037]
The high-speed binomial operation method of the compressed code string used in the present invention is based on the same principle as the conventional high-speed binomial logical operation method using a BDD. However, the present invention does not provide a table of all nodes of a BDD as input data in a state where it is stored in a storage device as in the related art, but sequentially reads compression-encoded code string data as an input. Simultaneously with the conventional method, the operation results are sequentially output at the same time.
[0038]
Further, in the present invention, during the process of compressing and decompressing a character string, when the capacity of the dictionary reaches a limit and a new character string x cannot be stored, among the stored character strings, A character string y that has never been used as a first half or a second half of another character string is searched for, and information representing the character string y (a character string and a registration number indicating a character string of the first half and the second half) is obtained. ) Is deleted, the information representing the new character string x is stored in the storage area pointed to by the same registration number as the character string y, and the subsequent processing can be continued. By deleting the information of the character string y, the chance of detecting the coincidence of the subsequences may be reduced and the compression rate may be reduced, but this does not cause the information of the character string data to be lost or distorted. Therefore, the arithmetic processing can be performed correctly.
[0039]
That is, in the present invention, even when the capacity of the storage device is not sufficient, it is possible to perform the arithmetic processing at the compression rate and the speed-up rate according to the capacity. On the other hand, in the conventional binary operation method using the BDD, since the node table of the BDD stored in the storage device is delivered as input data, large-scale data exceeding the capacity of the storage device is handled. It is impossible in principle.
[0040]
Furthermore, in the present invention, since the code string data which has been compression-encoded is sequentially read and the code string data of the operation result is sequentially output at the same time as the input, the processing is connected in series in a plurality of stages and the processing is executed in parallel in a pipeline manner. Can be. In the conventional binary operation processing method using the BDD, the processing of the next stage cannot be started unless the processing of the previous stage is completed. The present invention can be said to be superior to the conventional method in this respect as well.
[0041]
BEST MODE FOR CARRYING OUT THE INVENTION
1. String data compression encoding device:
First, the character string data compression encoding apparatus will be described.
FIG. 4 shows the configuration of the character string data compression encoding apparatus of the present invention.
The character string data compression encoding device 100 shown in FIG.ⁿ(Where n is an integer constant of n ≧ 0) is input as character string data, and code string data obtained by compressing the data without losing the information is output.
[0042]
The character string data compression encoding apparatus 100 includes an input processing unit 110 that sequentially reads character string data before compression from the beginning, and an output processing unit 120 that sequentially outputs code string data after compression and encoding from the beginning. , A compression encoding control unit 140 that controls them, and a decomposition rule 150.
In the character string data compression encoding apparatus 100, the length of the input character string data is just 2ⁿIt is assumed that there are (an integer with n ≧ 0). If character string data of any other length is input, the input processing unit 110 adds a plurality of special characters representing a blank to the end of the character string data and sends the character to the compression / encoding control unit 140. String data length is exactly 2ⁿ(Where n is an integer constant as small as possible).
[0043]
In the compression / encoding control unit 140, the inputprocessingThe character string data transferred from the unit 110 is decomposed into a number of subsequences and stored in the dictionary 130. The break position of the subsequence is based on a certain decomposition rule 150 regardless of the content of the character string data. That is, length 2ⁿIs divided into two substrings each having a half length of the first half and the second half of the input character string data, and the obtained substring is further divided into the first half and the second half respectively. 1,2,4,8,16, ..., 2 ⁿ Is decomposed into a number of subsequences. FIG. 5 shows an example in which character string data of the present invention is decomposed into sub-sequences. In the example shown in FIG. 5, a state in which character string data having a length of 8 is decomposed into sub-sequences is shown.
[0044]
The compression encoding control unit 140 sequentially reads the input character string data from the beginning based on the decomposition rule 50 described above, and reads the lengths 1, 2, 4, 8, 16, 2 ⁿ Is extracted, a different registration number is assigned to each different character string, and stored in the dictionary 130 as long as its capacity allows. By reading the input character string data while referring to the dictionary, it is detected that the same character string appears repeatedly in the character string data a plurality of times.
[0045]
2. String data recovery device:
Next, a character string data restoration device will be described.
FIG. 6 shows the configuration of the character string data restoration device of the present invention.
A character string data restoration device 200 shown in FIG. 2 restores code string data compressed and encoded by the character string data compression encoding device 100 to an original character string.
[0046]
The character string data decompression device 200 is used when the compressed character string data is compressed from the beginning, the output processing part 220 sequentially outputs the decompressed character string data from the beginning, and when the character string data is compressed. It is composed of a dictionary having a capacity equal to or greater than the dictionary 130 of the character string data compression encoding device 100 and a restoration control unit 240 for controlling them.
[0047]
When sequentially reading the compression-encoded code string data, the decompression control unit 240 outputs the character string as it is from the output processing unit 220 as it is when reading a partial string represented by a simple character string. When the set of the registration number of the subsequence and its character string is read, the subsequence is assigned the registration number and stored in the dictionary 230. When only the registration number of the subsequence is read, the registration is performed. The character string data is restored by extracting the subsequence indicated by the number from the dictionary 230 and outputting it.
[0048]
Further, when sequentially reading the compression-encoded code string data, the decompression control unit 240, when a character string represented by a set of two partial strings of the first half and the second half is input, the first half Then, the character string is represented by a set of the registration numbers of the character strings in the latter half and stored in the dictionary 230. As a result, a dictionary having exactly the same contents as the dictionary 130 created internally when compression encoding is performed by the above-described character string data compression encoding apparatus 100 is restored.
[0049]
Further, when extracting and outputting the partial sequence stored in the dictionary 230, the restoration control unit 240 sequentially refers to the first half and the second half of the partial sequence from the dictionary 230, and further extracts the first half from the partial sequence. In the case where the partial sequence is expressed as a set of the partial sequence and the latter half, the sequence is further referred to, a subsequence of length 1 is finally extracted and output, and the entire subsequence is output.
[0050]
3. Character string data processing unit:
Next, the character string data processing device will be described.
FIG. 7 shows the configuration of the character string data processing device of the present invention.
The character string data arithmetic processing device shown in FIG. 11 includes two compression code string decompression devices (the same configuration as the above-described character string data decompression device) as input processing units 310 and 320, and compression codes as output processing units 360. , An arithmetic processing unit 330 for performing a binomial operation, an arithmetic high-speed dictionary 350 having a finite capacity, and a control unit 340 for controlling them. Become.
[0051]
This device calculates and outputs one character for any two characters, and performs a certain binary operation (for example, a logical operation such as a logical product and a logical sum or an arithmetic operation such as addition, subtraction, multiplication, and division). Is defined, two code string data compressed in the above-described encoding method are sequentially read, and two character string data that would be obtained by restoring the compressed code string are described below. One character string data representing the result of calculating a predetermined binary operation is calculated for characters at the same position from the beginning, and this is sequentially output in a state compressed by the above-described encoding method. is there.
[0052]
In the two input processing units 310 and 320, the compressed code string data is sequentially read from the beginning, decompressed into character string data and transferred to the arithmetic processing unit 330. The two character string data transferred are sequentially calculated by performing a binary operation between characters at the same position from the beginning, and the calculation result is transferred to the output processing unit 360 as new character string data. In the output processing unit 36, the character string data transferred from the arithmetic processing unit 330 is compression-encoded and output as code string data.
[0053]
However, simply restoring the character string data, sequentially calculating the data from the beginning, and compressing the data again requires a calculation time proportional to the length of the restored character string data, which is not efficient. Therefore, the present invention includes means for speeding up the arithmetic processing as described below.
The two input processing units 310 and 320 and the output processing unit 360 periodically report the information of their registration numbers to the control unit 340 when the subsequences are stored in or referenced from the respective dictionaries.
[0054]
The control unit 340 always monitors which subsequence corresponds to the result of the binomial operation on a certain set of two input subsequences, and every time a set of registration numbers of these three subsequences is obtained, By storing the data in the speed-up dictionary 350 as much as the capacity permits and sequentially performing calculations while referring to the calculation speed-up dictionary 350, it is detected that the same set of input subsequences appears a plurality of times.
[0055]
Further, when the control unit 340 detects the same set of input subsequences, the input processing units 310 and 320 omit the process of restoring the registration number of the subsequence into character string data from the second time on. The control command is sent, and the output processing unit 360 sends the control command so as to directly output the registration number of the partial sequence of the calculation result stored in the calculation acceleration dictionary 350.
[0056]
According to the above method, when a large number of sets of the same input subsequence appear, the entire processing time can be reduced as compared with a method of simply restoring, calculating, and recompressing. The rate of the speed-up is determined by the frequency of appearance of the same set of input sub-sequences. When the capacity of the arithmetic processing speed-up dictionary 350 is sufficient, the high-speed rate almost proportional to the compression ratio of the input and output code string data is obtained. Conversion rate is obtained.
[0057]
【Example】
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[First Embodiment]
A first embodiment will be described based on a character string data compression encoding apparatus 100 shown in FIG.
[0058]
FIG. 8 is a diagram for explaining the operation of the character string data compression encoding apparatus according to the first embodiment of the present invention, and illustrates the operation when eight character string data having a length of “ABABABAA” is input. Here is an example.
FIG. 9 shows an example of a dictionary in which substrings of character string data according to the first embodiment of the present invention are registered. Hereinafter, a description will be given based on the examples of FIGS. 8 and 9.
[0059]
The character string is read by the input processing unit 110 character by character from the left. First, a character of length "A" is recognized. Next, a character of length "B" is recognized. Here, since a partial sequence "AB" appears for the first time, it is registered in the dictionary 130 with a registration number of partial sequence 1 given thereto. The dictionary stores a character string of length 2 in which the characters "A" and "B" are linked.
[0060]
Next, when the third character and the fourth character are read, the substring "AB" is recognized again. Since this is the same character string as substring 1 already registered, there is no need to newly register it.
Next, the partial sequence “ABAB” from the first character to the fourth character is recognized. Since this is a new character string, it is registered in the dictionary 130 with a registration number of subsequence 2 added. Since the first half “AB” and the second half “AB” of this character string are the same character string as the partial string 1, respectively, the dictionary 130 stores the partial string 1 as a partial string having a length of 4 by linking two partial strings 1. .
[0061]
Subsequently, the fifth character and the sixth character are read, and the substring "AB" is recognized. However, since this is the same character string as the substring 1, there is no need to newly register it.
Next, the seventh and eighth characters are read, and a substring "AA" is recognized. Since this is a newly appearing character string, it is registered in the dictionary 130 with a registration number of subsequence 3 added. The dictionary 130 is stored as a substring of length 2 in which two characters “A” are connected.
[0062]
Next, the character string “ABAA” from the fourth character to the eighth character is recognized. Since this is also a character string that appears for the first time, a registration number of subsequence 4 is given and registered in the dictionary 130. The dictionary 130 stores the subsequence 1 and the subsequence 3 as a subsequence having a length of four.
Finally, the partial sequence from the first character to the eighth character is recognized, and this is registered in the dictionary 130 with the registration number of the partial sequence 5 assigned. The dictionary 130 is stored as a substring having a length of 8 in which the subsequences 2 and 4 are connected.
[0063]
As described above, the input character data is decomposed into substrings and stored in the dictionary 130.
Next, how the input data read as described above is compression-encoded and output will be described. FIG. 10 shows an example of code string data obtained by compressing and encoding character string data according to the first embodiment of the present invention. The code string data shown in FIG. 8 is obtained by compressing and encoding the character string data shown in FIG.
[0064]
When a substring consisting of the same character string appears repeatedly in the input character string data a plurality of times, the compression encoding control unit 140 outputs the information of the character string and its registration number the first time, and the second and subsequent times Compresses and encodes data by outputting only the registration number indicating the subsequence. In FIG. 10, first, the characters “A” and “B” are output as they are, and then “AB” is registered with the registration number “subsequence 1” by the code “(column 1 = A, B)”. Output that. Thereafter, instead of outputting the character string "AB", only the code "column 1" is output to compress the data.
[0065]
Regarding the timing of outputting the compressed code string data, the output is performed simultaneously and in parallel with the input each time a new partial string is recognized and registered without waiting for the input character string data to be completely read.
The length of the code string data obtained by performing such compression encoding is substantially proportional to the number of substrings newly appearing in the input character string data. Therefore, when input character string data in which the same subsequence appears many times is given, a very high compression ratio can be obtained.
[0066]
The calculation time for registering and referring to the dictionary 130 is determined by a hash table technique, regardless of the number of registered data in the dictionary, and is calculated once in a certain constant time, assuming a storage device that can be randomly accessed. -Reference processing can be executed. Therefore, the total calculation time is almost proportional to the length of the input character string data.
By the way, in an actual device, it is impossible to realize a dictionary having an infinite capacity, and a dictionary 130 having a certain finite capacity is used. Therefore, it is necessary to take into consideration that when large-scale character string data is input, the dictionary 130 becomes full and a new subsequence cannot be stored.
[0067]
Therefore, in the present embodiment, when the dictionary 130 is full, a partial sequence that has not been used as the former half or the latter half of another partial sequence is searched for in the partial sequences stored so far. , The information representing the subsequence (a set of registration numbers indicating the first half and the second half of the character string) is temporarily deleted, and a new subsequence is stored in the storage area indicated by the same registration number as the deleted subsequence. The information to be stored is stored, and the subsequent processing proceeds (hereinafter, this processing is referred to as “dictionary overflow continuation processing”).
[0068]
By performing the above dictionary overflow continuation processing, a part of the information of the past partial sequence is lost, and there is a possibility that the partial sequence which should have been able to be detected and compressed by the coincidence cannot be detected. Since the character string that could not be compressed is output as it is, the compression ratio may be reduced, but there is no fear that information is lost or distorted. If the decompression device has a dictionary having a capacity equal to or larger than that of the compression device, it is guaranteed that the data can be correctly reconstructed and restored.
[0069]
Further, in the actual character string data, if it is assumed that a subsequence located at a position closer to the distance is more likely to appear more than once, as shown in the above dictionary overflow continuation processing, a subsequence that is used less frequently is used. It is conceivable that erasing and replacing with the latest subsequence would make more efficient use of the limited dictionary capacity, and as a result, a reduction in the compression ratio would be reduced.
[0070]
When performing the dictionary overflow continuation process, in order to quickly find a subsequence that may be deleted, a counter that counts the number of times a certain subsequence is referred to by other subsequences is provided for each subsequence information. When a new reference is made or a reference is lost, the value of this counter is increased or decreased. By always listing the subsequences in which the counter value is 0, the dictionary 130 can be updated immediately when necessary.
[0071]
[Second embodiment]
The following second embodiment will be described based on the character string data restoration device 200 shown in FIG.
In the character string data restoration apparatus 200, when the compression-encoded code string data shown in FIG. 10 in the first embodiment is input, the dictionary shown in FIG. 9 is restored.
[0072]
For example, in FIG. 9, in order to extract and output a subsequence stored in the dictionary 230 as the subsequence 4, the restoration control unit 240 first refers to the subsequence 1 which is the first half of the subsequence 4, Further, "A" is output with reference to the first half of the partial sequence 1, and then "B" is output with reference to the second half of the partial sequence 1. This completes the output of the partial sequence 1.
[0073]
Next, the process returns to the sub-sequence 4 once, refers to the sub-sequence 3 which is the latter half, and outputs “A” and “A” in the same manner. In this way, the entire partial sequence 4 can be output.
Regarding the timing of outputting the decompressed character string data, the output is performed simultaneously and in parallel each time a partial string is extracted from the input processing unit 210 or the dictionary 230 without waiting for the compressed code string data to be read to the end. .
[0074]
[Third embodiment]
As a third embodiment, a character string data processing device shown in FIG. 7 will be described. The character string data processing device performs a binomial operation between the compressed code sequence data. The compressed code sequence decompression device 200 of the second embodiment is used as the input processing units 310 and 320. The character string data compression / encoding device 100 according to the first embodiment is defined as an output processing unit 360, and an arithmetic processing unit 330 that performs a binary operation on two restored character string data input from the input processing units 310 and 320. And a control unit 340 for controlling each of these components.
[0075]
In the control unit 340, when the two input processing units 310 and 320 and the output processing unit 360 store or refer to the respective sub-strings in the respective dictionaries (corresponding to the dictionary 230 in FIG. 6), the respective registration numbers And monitors which subsequence corresponds to the result of the binomial operation in the arithmetic processing unit 330 for a certain set of two input subsequences. Each time a set of registration numbers of three subsequences of the corresponding subsequence is obtained, it is stored in the arithmetic processing acceleration dictionary 350. With reference to the arithmetic processing speed-up dictionary 350, it is detected that the same set of input subsequences appears a plurality of times.
[0076]
Here, when the same set of input subsequences is detected, the second and subsequent times are instructed to the input processing units 310 and 320 to omit the process of restoring the registration number of the subsequence to character string data. And the output processing unit 360 issues a command to output the registration number of the partial sequence of the operation result stored in the operation processing acceleration dictionary 350 as it is. Thereby, since the same arithmetic processing is not repeated, the speed can be increased.
[0077]
FIG. 11 shows an example of an arithmetic processing acceleration dictionary according to the third embodiment of the present invention. The registration number of three substrings is registered as one set in the arithmetic processing acceleration dictionary 350 including two inputs and one arithmetic result. Assuming a storage device that can be accessed randomly, the hash table technique allows one registration / reference in a certain constant time without depending on the number of entries in the dictionary 350.
[0078]
In an actual device, the capacity of the arithmetic processing acceleration dictionary 350 is finite, but when the dictionary 350 becomes full, old information is appropriately deleted and new information is overwritten. As a result, an opportunity to detect a subsequence that can be omitted may be lost, and the processing speed may be reduced. However, even in such a case, the content of the code string data of the operation result is not affected only by lowering the speed.
[0079]
It should be noted that the present invention is not limited to the above-described embodiment, but can be variously modified and applied within the scope of the claims.
[0080]
【The invention's effect】
As described above, according to the present invention, the compression-encodedWithout necessarily restoring completelyIt is possible to provide an apparatus for directly performing an arithmetic processing. As a result, the processing speed can be increased substantially in proportion to the data compression ratio.
Further, according to the present invention, even when all the input data cannot be stored in the storage device in advance, it is possible to sequentially execute the arithmetic processing from the beginning of the input compressed code string data and sequentially output the arithmetic results. It is possible. As a result, arithmetic processing of large-scale data that cannot be handled by the conventional method can be performed.
[0081]
Further, according to the present invention, by connecting a plurality of arithmetic processing units in series, it is possible to increase the processing speed by executing the processing in parallel in a pipeline manner.
[Brief description of the drawings]
FIG. 1 is a block diagram of the principle (character string data compression encoding apparatus) of the present invention.
FIG. 2 is a block diagram of the principle (character string data restoration device) of the present invention.
FIG. 3 is a diagram illustrating the principle configuration (character string data processing device) of the present invention.
FIG. 4 is a configuration diagram of a character string data compression encoding device of the present invention.
FIG. 5 is a diagram for explaining a subsequence decomposition rule according to the present invention.
FIG. 6 is a configuration diagram of a character string data restoration device of the present invention.
FIG. 7 is a configuration diagram of a character string data processing device of the present invention.
FIG. 8 is a diagram for explaining the operation of the character string data compression encoding apparatus according to the first embodiment of the present invention.
FIG. 9 is an example of a dictionary in which substrings of character string data according to the first embodiment of the present invention are registered.
FIG. 10 is an example of code string data obtained by compressing and encoding character string data according to the first embodiment of the present invention.
FIG. 11 is an example of an arithmetic processing acceleration dictionary according to a third embodiment of the present invention.
FIG. 12₁, X₂, X₃3 is an example of a truth table expressing logic function data having the three variables as inputs.
FIG. 13 is an example of a binary tree-like graph obtained by substituting 0 and 1 for input variables with respect to logic function data.
FIG. 14 is a diagram illustrating a process of sharing a BDD.
FIG. 15 is a diagram for explaining a process of making a BDD non-redundant.
FIG. 16 is an example of a BDD representing logic function data.
FIG. 17 is an example of a table listing nodes of the BDD in order to store the BDD in a storage device.
[Explanation of symbols]
100 Character string data compression encoding device
110 input processing means, input processing unit
120 output processing means, output processing unit
130 dictionaries
140 compression encoding control unit
141 detection means
142 output control means
150 Decomposition rules
200 character string data recovery device
210 input processing means, input processing unit
220 output processing means, output processing unit
230 dictionaries
240 restoration control means, restoration control unit
300 Character string data processing unit
310, 320 Input processing means
330 arithmetic processing means
340 control unit
350 High Speed Processing Dictionary
360 output processing means

Claims (8)

A character string data compression / encoding device that receives character string data having a length of 2 ⁿ (where n is an integer constant of 0 or more) and outputs code string data obtained by compressing the data without losing the information. So,
Input processing means for sequentially reading the character string data before compression from the beginning,
Output processing means for sequentially outputting code string data after compression encoding from the beginning,
A finite capacity dictionary,
The separated length 2 ⁿ input character string data input from the input processing means into two subsequences of each half length of the first half and the second half portion, the first half of the resulting the portion segment sequence and the latter half portion to separate, by repeating this, the length 1,2,4,8,16, decomposition rule to be referenced in order to decompose the subsequence of ... ^{2 n,}
While sequentially reads the input character string data input from said input processing means with reference to the decomposition rule, the length 1,2,4,8,6, ..., remove subsequences of 2 ^n, different By assigning a different registration number to each character string, storing it in the dictionary as long as its capacity permits, and reading the input character string data while referring to the dictionary, the same character string is input into the input character string data. Detecting means for detecting that the character string has repeatedly appeared plural times;
When a substring consisting of the same character string appears repeatedly in the input character string data a plurality of times, the output processing unit is instructed to output a set of a character string and a registration number of the character string at the first time. And a second and subsequent output control means for instructing the output processing means to output only the registration number indicating the subsequence. When character string data of an arbitrary length is input, a special character representing a blank character string is used so that the length of the input character string data is exactly 2 ⁿ (where n is an integer constant as small as possible). 2. The character string data compression encoding apparatus according to claim 1, further comprising means for adding a plurality of data to the end of the data. The detecting means,
An arbitrary character string a having a length of 2 or more is decomposed into two partial strings a1 and a2 each having a half length of the first half and the second half, and the character strings of the partial strings a1 and a2 are If the character string a is already stored in the dictionary, the character string a is represented by a pair of the registration numbers indicating the character strings of the subsequence a1 and the subsequence a2. Storage means for giving a registration number and storing it in the dictionary,
The output control means,
When a character string a having a length of 2 or more is output, if the character strings of the two partial strings a1 and a2 of the first half and the second half are not stored in the dictionary, the partial strings a1 and The character string a is expressed by sequentially outputting the character strings of the column a2, and when one or both of the character strings of the subsequence a1 and the subsequence a2 are stored in the dictionary, the stored character 2. The character string data compression encoding apparatus according to claim 1, further comprising means for outputting a registration number indicated by the character string instead of outputting the string. The storage means,
When the capacity of the dictionary reaches the limit and a new character string x cannot be stored, it is still used as a first half or a second half of another character string in the character strings stored so far. Search for a character string y that has not been deleted, and temporarily delete a set of registration numbers indicating the first half and the second half of the character string y representing the character string y, and store the storage pointed by the same registration number as the deleted character string y. 4. The character string data compression encoding apparatus according to claim 3, further comprising means for storing information representing a new character string x in an area and proceeding with subsequent processing. A character string data decompression device for decompressing and extracting original character string data from code sequence data compressed and encoded by a character string data compression encoding device,
Input processing means for sequentially reading compression-encoded code string data from the beginning,
Output processing means for sequentially outputting the restored character string data from the beginning,
A dictionary equivalent to or larger than the dictionary of the character string data compression encoding device used when compressing the character string data,
When sequentially reading the compression-encoded code string data obtained from the input processing means, if a partial string represented as a simple character string is read, the output processing is performed such that the character string is output as it is. Instructing the means, when reading the set of the registration number and the character string of the subsequence, storing the character string with the registration number added to the dictionary and reading only the registration number of the subsequence And a restoration control unit for instructing the output processing unit to take out a partial string indicated by the registration number from the dictionary and output the extracted partial string. The restoration control means includes:
When sequentially reading the compression-encoded code string data, when a character string represented by a set of two partial strings of a first half and a second half is input, the partial string of the first half and the second half is input. Means for expressing the character string with a set of registration numbers and storing it in the dictionary;
When extracting and outputting the partial sequence stored in the dictionary, the partial sequence of the first half and the second half of the partial sequence is sequentially referred to from the dictionary, and those partial sequences are further divided into the first half and the second half. 6. The character string data restoring device according to claim 5, further comprising means for, when represented as a set, sequentially referring to them and extracting and outputting a substring having a length of one. Calculates and outputs one character for any two characters. If a certain binary operation is defined, two character string data of the same length are input and the same from the beginning. A character string data processing device that outputs one character string data representing a result of calculating a predetermined binary operation for characters at positions,
First input means for sequentially reading the compression-encoded code string data from the beginning, first output means for sequentially outputting the restored character string data from the beginning, and character string data used when the character string data is compressed When sequentially reading the first dictionary having a capacity equal to or larger than the dictionary of the compression encoding device and the compression-encoded code string data acquired from the first input means, a short simple character string is used. When the expressed substring is read, the first output means is instructed to output the character string as it is, and when the set of the registration number and the character string of the substring is read, the When the character string is given the registration number in the first dictionary and stored, and only the registration number of the subsequence is read, the subsequence indicated by the registration number is taken out from the first dictionary and output. To the first output means Two consisted string data restoration device, two input processing means for transferring to restore each input code string data to the character string data and a restoration control unit that,
Arithmetic processing means for sequentially calculating a binary operation between characters at the same position from the beginning of the two character string data transferred from the input processing means, and transferring the result as new character string data;
A second input unit for sequentially reading the character string data before compression from the beginning, a second output unit for sequentially outputting the code string data after compression and encoding from the beginning, a second dictionary having a certain finite capacity, The input character string data having a length of ²ⁿ input from the second input means is divided into two substrings each having a half length of the first half and the second half thereof, and the first half of the obtained partial string is divided into two substrings. Separate the rear half portion, by repeating this, the length 1,2,4,8,16, and decomposition rules referred to decompose the subsequence of ... 2 ^n, said second referring to the decomposition rules , While sequentially reading the input character string data input from the input means, extracts the sub-sequences of lengths 1, 2, 4, 8, 16,... ²ⁿ and assigns a different registration number to each different character string Then, the data is stored in the second dictionary as much as the capacity allows, and the second dictionary is referred to. However, by reading the input character string data, detecting means for detecting that the same character string repeatedly appears in the input character string data a plurality of times, and detecting the input character string data from the same character string. If the sub-sequence appears repeatedly a plurality of times, the second output unit is instructed to output a set of a character string and a registration number of the character string at the first time. An output control means for instructing the second output means to output only a registration number indicating a column, wherein the character string data transferred from the arithmetic processing means is compressed and encoded. Output processing means for converting the data into code string data and outputting the code string data. The input processing unit and the output processing unit further include a transfer unit that transfers information of each registration number when a subsequence is stored or referred to in each of the first dictionary or the second dictionary. And
The information of the registration number is obtained from the transfer means, and the subsequence corresponding to the result of the binomial operation on a certain set of two input subsequences is determined by the three input subsequences of two input subsequences and one output subsequence. As a set of registration numbers of two subsequences, the same input subsequence is stored in an operation acceleration dictionary having a finite capacity as long as the capacity permits, and by sequentially performing operation processing while referring to the operation acceleration dictionary. Detecting means for detecting that the set of characters has appeared a plurality of times;
When the same set of input sub-sequences appears, the input processing means controls so as to omit the process of restoring the registration number of the sub-sequence to character string data, and the output processing means stores in the arithmetic high-speed processing dictionary. 8. The character string data processing apparatus according to claim 7, further comprising control means for controlling so as to output the registration number of the stored partial sequence of the operation result as it is.

Images