JPH11317671A - Data compressing device and data compressing method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data compressing device which accelerates data compression processing and a data compressing method. SOLUTION: A test pattern compressing device 10 includes a BW conversion application deciding part 11, a BW converting part 12, a comparing part 13, a data selecting part 14 and a run length compressing part 15. The part 11 measures the transition number of times of input data, decides to the effect that Burrows-Wheeler(BW) conversion processing is applied to the input data, when the measurement data is a prescribed value or less or decides to the effect that the conversion processing is not applied, when it is larger than the prescribed value. When the conversion processing is not applied, a conversion operation is not performed by the part 12 but run-length-coding processing is performed to the input data by the part 15.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data compression apparatus and a data compression method for compressing test pattern data and the like used in an integrated circuit test.

[0002]

2. Description of the Related Art Conventionally, Burrows-Wh is used as a method for compressing a large amount of test pattern data for testing an integrated circuit.
A method of compressing data by combining eeler transform processing and run-length encoding processing is known.

[0003] The Burrows-Wheeler transform (hereinafter referred to as "BW transform") is described in, for example, Burrows M. and Whee.
ler D., “A Block-sorting Lossless Data Compression
Algorithm, ”SRC Research Report 124, Digital Syste
ms Research Center, Palo Alto, CA, May 1994. In the BW conversion, a rearrangement operation for collecting symbols of the same context or similar context included in input data is performed. Here, "context" represents a data string arbitrarily extracted from input data. For example,
When the input data is character data, the context represents a character string arbitrarily extracted from the input character string. When the input data sequence is sufficiently long, the same symbols are gathered by the BW conversion, and the number of transitions of the BW-converted data is smaller than the number of transitions of the original input data. Here, “the number of transitions” means that two adjacent data in the data string are 0
Represents the number of changes from 1, 1 to X. For this reason, it is possible to efficiently compress the BW-transformed data by combining it with run-length encoding.

More specifically, the input data is BW-converted, the number of transitions of the original input data is compared with the number of transitions of the data after the BW conversion, and the run-length code is applied to data having a smaller number of transitions. The compression of the test pattern data is performed by using the conversion.

[0005]

By the way, in the above-mentioned conventional data compression method, it is impossible to accurately know the number of transitions after the BW conversion of certain data.
The input data is once BW-converted to check whether the number of transitions is reduced by the conversion, but the processing speed of the BW conversion is very low, so that much time was required for the data compression processing. .

In particular, as the degree of integration of an integrated circuit increases, the data size of a test pattern for testing the integrated circuit has been increasing year by year. ASIC for integrated circuits
(Application Specific Integrated Circuit), the number of integrated circuits has been increased in a small number and variety, and accordingly, the types of test patterns have also increased. Therefore, in order to compress many test patterns in a practical processing time, a faster test pattern compression method has been desired.

[0007] The present invention has been made in view of the above points, and an object of the present invention is to provide a data compression apparatus and a data compression method capable of speeding up data compression processing.

[0008]

In order to solve the above-mentioned problems, a data compression apparatus and a data compression method according to the present invention perform a rearrangement process for collecting symbols of the same or similar context contained in input data. It is determined in advance whether or not the number of transitions of the data decreases when the data is not actually rearranged. If it is determined that the number of transitions does not decrease, the rearrangement process is stopped. Therefore, the number of input data to be rearranged can be reduced, and the speed of data compression can be increased.

In particular, the above-described rearrangement operation is performed in Burrows-Wh
It is preferably performed by eeler conversion processing. In addition, it is preferable that test pattern data for an integrated circuit test be compressed. Since the Burrows-Wheeler transform takes a lot of time to rearrange the data sequence, this rearrangement process is not performed for all input data, but the number of data transitions decreases when the rearrangement process is performed. For input data that is not used, the effect of reducing the processing time by stopping the rearrangement process is significant. Further, the test pattern for the integrated circuit test is a huge amount of data in which bit data “1” and “0” are combined, and if the same bit data can be continuously arranged by rearranging, it is executed later. Combined with the run-length coding described above, the data amount can be significantly reduced by compression.

Further, it is preferable that the above-described determination operation as to whether or not the number of transitions of data is reduced is performed based on the number of transitions of input data. If the number of transitions of the input data is small, it is not expected that the number of transitions will be further reduced even if the rearrangement process is performed. And the speed of the data compression process can be increased by the amount corresponding to the stop of the rearrangement operation.

The above-described determination operation as to whether or not the number of data transitions is reduced is performed by a rearrangement operation of restricting the length of a context to input data and collecting symbols of the same or similar context. For example, it is preferable to perform the operation based on the result of performing the data rearrangement operation by the S conversion process. Since there is a correlation between the number of transitions of the data obtained by the rearrangement process that limits the length of the context and the number of transitions of the data obtained by the rearrangement process that does not limit the length of the context, a context that enables high-speed processing By performing the sorting process that limits the length of the context, it is possible to determine whether or not the sorting process that does not limit the length of the context reduces the number of data transitions. By appropriately stopping the rearrangement processing, unnecessary rearrangement processing is almost exactly eliminated, and the speed of the data compression processing can be significantly increased.

It is preferable that the above-described determination operation as to whether or not the number of data transitions is reduced is performed based on the value of the entropy of the input data. The entropy of the input data is obtained by averaging the probability of occurrence of each symbol in the input data for all symbols and adding a negative sign. The magnitude of this entropy and the input data are rearranged. Since there is a correlation with the number of subsequent data transitions, it is possible to determine whether or not the data rearrangement reduces the number of data transitions by measuring this entropy, and determine the length of the context according to the determination result. By appropriately stopping the rearrangement processing that does not limit the data size, unnecessary data rearrangement processing can be almost exactly eliminated, and the data compression processing can be significantly speeded up.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A test pattern compression apparatus according to an embodiment of the present invention determines whether or not to apply BW conversion to input test pattern data by using high-speed means. The feature is that the test pattern to be compressed is compressed at high speed by stopping the application of the BW conversion to the input data whose transition number is not reduced by the conversion. Hereinafter, details of a test pattern compression apparatus according to an embodiment to which the present invention is applied will be described with reference to the drawings.

[First Embodiment] FIG. 1 is a diagram showing a configuration of a test pattern compression apparatus according to a first embodiment of the present invention. As shown in the figure, the test pattern compression apparatus 10 of the present embodiment includes a BW conversion
It comprises a W conversion unit 12, a comparison unit 13, a data selection unit 14, and a run-length compression unit 15.

The BW conversion application determining unit 11 measures the number of transitions φ _IN of the input data, and based on the number of transitions φ _IN ,
It is determined whether to apply the BW conversion to the input data.
Specifically, when the number of transitions φ _IN is equal to or less than a predetermined value, it is determined that the BW conversion is to be applied to the input data, and 1 is set to the flag f1. Conversely, when the number of transitions φ _IN is larger than the predetermined value, it is determined that the BW conversion is not applied to the input data, and 0 is set in the flag f1. For example, considering the case where the predetermined value is set to 1, if the number of transitions φ _IN of the input data is equal to or less than the predetermined value, that is, 0 or 1, the number of transitions is already the lower limit, and the transition is performed by BW conversion. We cannot hope to reduce the number of times. Therefore, it is not necessary to perform the BW conversion on the input data having such a transition number, and it is determined that the BW conversion is not performed.

[0016] The BW conversion unit 12 is a BW conversion application determination unit 1
When it is determined that the BW conversion is to be applied according to 1, the BW conversion is performed on the input data, and the number of transitions φ _BW of the data after the BW conversion is measured. The comparison unit 13 compares the number of transitions φ _IN of the input data obtained by the BW conversion application determination unit 11 with the number of transitions φ _BW of the data after BW conversion obtained by the BW conversion unit 12.
If φ _IN > φ _BW , 1 is set to the flag f2, and φ
0 in the flag f2 is set when the _IN ≦ _{φ BW.}

The data selector 14 receives the input data and the data obtained by performing the BW conversion on the input data.
Based on the comparison result of step 3, data with a smaller number of transitions is selected. Specifically, when the flag f2 set by the comparing unit 13 is 1, the BW-converted data is selected, and when the flag f2 is 0, the input data is selected. The run-length compression unit 15 performs data compression on the data selected by the data selection unit 14 by run-length encoding.

The above-described BW conversion application determining unit 11 corresponds to a sorting operation application determining unit, the BW converting unit 12 corresponds to a first sorting unit, and the data selecting unit 14 corresponds to a data selecting unit.

Next, the operation when the test pattern is compressed by using the test pattern compression apparatus 10 of the present embodiment will be described. The test pattern compression apparatus 10 performs various processes using data of a predetermined size L as one input unit, and data larger than L is divided into a plurality of sub-data of size L in advance, and the divided data is divided. Each sub data is input to the test pan compression apparatus 10.

FIG. 2 is a flowchart showing an operation procedure of a test pattern compression operation performed in the test pattern compression apparatus 10 of the present embodiment. First, the BW conversion application determining unit 11 measures the number of transitions m of the input test pattern data (Step 101), and
The application of the W conversion is determined (step 102). When the BW conversion is applied to the input data, 1 is set to the flag f1, and when not applied, 0 is set to the flag f1.

Next, the BW converter 12 sets the flag f1 to 1
Is determined (step 103). Flag f
If 1 is 1, the BW conversion unit 12 performs BW conversion processing on the input data (step 104),
The number of transitions n of the W-converted data is measured, and the measured value is set to φ _BW (step 105). On the other hand, when the flag f1 is 0, the BW conversion unit 12 sets the input data size L, which is the upper limit value of the number of transitions, to φ _BW (step 106).

Next, the comparing section 13 compares the values of φ _IN and φ _BW and sets a flag f2 based on the comparison result (step 107). Specifically, the comparison unit 13 determines that φ _IN ≦ φ
_In the case of _BW , 0 is set to the flag f2, and in the case of φ _IN > φ _BW , 1 is set to the flag f2. Next, the data selection unit 14
It is determined whether the flag f2 is 1 (step 10).
8) If the flag f2 is 1, the BW-converted data is selected (step 109), and if the flag f2 is 0, the input data is selected (step 110).

Next, the run-length compression unit 15 performs a compression process using run-length coding on the data selected by the data selection unit 14 (step 11).
1) The compressed data is output (step 112), and a series of data compression processing ends. In order to decompress the compressed data to the original test pattern data, the compressed data output from the run-length compression unit 15 includes a check whether the run-length encoded compressed data is BW-converted or not. A flag f2 indicating whether or not the flag is added.

As described above, the data compressed by the test pattern compression apparatus 10 of the present embodiment is subjected to BW inverse conversion after run-length decoding if the flag f2 added to the compressed data is 1. Thus, the original test pattern data can be completely expanded to the original test pattern data. When the flag f2 is 0, only the run-length decoding can be performed to completely expand the original test pattern data.

Next, BW conversion processing and run-length encoding processing performed at the time of data compression, and run-length decoding processing and inverse BW conversion processing performed at the time of data expansion will be described based on specific examples.

FIG. 3 is a flowchart showing the operation procedure of the BW conversion process performed in the BW conversion unit 12. First, B
The W conversion unit 12 acquires the input character string S and sets the size (length) of the input character string to n (Step 201).
Hereinafter, for example, the description will be made using “abraca” as the input character string S of 6 characters where n = 6.

Next, the BW conversion unit 12 generates n cyclic character strings (cyclically shifted character strings) corresponding to the obtained input character string S (step 202). As shown in FIG. 4A, the cyclic character string is a character string in which the first character of the character string is cyclically moved to the end.
In the example of the character string S ('abraca') used here,
As shown in FIG. 4B, six cyclic character strings are obtained.

Next, the BW conversion unit 12 determines in step 202
The matrix M is created by rearranging the six cyclic character strings obtained in (1) in the order registered in the dictionary (step 203). S =
In the example of 'abraca', the data is rearranged as shown in FIG. 5, such as 'aa', then 'ab', and then 'ac'.

Next, the BW conversion unit 12 obtains, based on the created matrix M, a final character string L obtained from the last column of the matrix M and a row number I equal to the input character string S (step 204). , 205). In the example of S = 'abraca', as shown in FIG. 6, L = 'caraab' and I = 2
Is obtained.

As described above, the character string S = 'abraca' is replaced with the final character string L = 'caraab' and the line number I
= 2. Finally, the BW converter 1
2 outputs the BW-converted data (step 2).
06), the process ends.

FIG. 7 is a flowchart showing the operation procedure of the run-length encoding process performed in the run-length compression section 15. First, the run-length compression unit 15 obtains one-character data from an input character string and sets the obtained data to a channel (step 301), and sets the obtained character ch to an initial value initial of a symbol for generating a run-length code. Then, the count i is reset to 0 (step 30).
2).

Next, the run-length compression unit 15 obtains the next character from the input data, sets it to ch (step 303), and increments the number of characters by adding 1 to the count i (step 304). ). Further, the run-length compression unit 15 stores the acquired character ch and the initial value initi.
It is determined whether or not al matches (step 30).
5) If they match, it is determined whether or not input data remains (step 306). If there is any remaining, the processing of step 303 and thereafter is repeated.

When the character ch does not match the initial value initial (when a negative determination is made in step 305) or when no input data remains (when a negative determination is made in step 306), Next, the run-length compression unit 15 sets the initial value initi
An encoding process is performed using al and the obtained count number i (step 307). Then, the run length compression unit 1
5 determines whether or not input data remains (step 308). If input data remains, the processing from step 302 onward is repeated. If no input data remains, compressed data remains. (Step 3
09), the process ends.

FIG. 8 is a flowchart showing the operation procedure of the run-length decoding process. For example, the operation procedure in the case where the run-length code is expanded by a run-length decoding unit (not shown) to obtain the original character string is shown. ing. First, the run-length decoding unit obtains one run-length code from the compressed data (step 401), and obtains an initial value initial and the number n of consecutive characters from the run-length code (step 402). Next, the run-length decoding unit:
After the initial value "initial" is copied by the number of consecutive times n (step 403), it is determined whether or not the run-length code remains (step 404). If the run-length code remains, the process from step 401 onward is repeated. Also,
If no run-length code remains, the run-length decoding unit outputs the decompressed data (step 405), and ends the process.

FIG. 9 is a flowchart showing the operation procedure of the inverse BW conversion process. For example, the operation procedure in the case where an inverse BW conversion unit (not shown) performs the inverse BW conversion is shown. In the following, the final character string L =
The description will be made using an example of 'carab', line number I = 2. First, the inverse BW conversion unit acquires the converted data, and sets the final character string L, the line number I, and the character string length n obtained from the converted data (step 501).
In the above example, L = 'carab', I = 2, n =
6 is set.

Next, the inverse BW conversion unit rearranges the characters of the obtained final character string L to generate a character string F (step 502). In the example of L = 'carab', 'aaaabcr' is generated as the character string F. Further, the inverse BW conversion unit calculates the i-th character L of the final character string L and the character string F.
A matrix T indicating the correspondence between [i] and F [i] is created (step 503). Here, T is L [T [i]] = F [i]
Is a matrix that satisfies Also, if the same character ch is used in the final character string L, they are associated with each other so that the channels in the final character string L are in the same order as the channels in the character string F. Therefore, L = 'cara
In the example where ab 'and F =' aaaabcr ', as shown in FIG. 10, T = [245613] is obtained.

Next, the inverse BW converter sets the count number i to 0.
(Step 504), 1 is added to the count number i (step 505), and the character string F, line number I,
The i-th character S [i] of the original input character string S is restored from the matrix T and output (steps 506 and 507). Here, S [i] is S [i] = F [T ⁱ⁻¹ [I]], T [i]
⁰ [I] = I, T ^{i + 1} [I] = T [T ⁱ [I]].

Next, the inverse BW conversion unit determines whether there is a character that has not been restored in the input character string S, that is, whether i <n is satisfied (step 508). If it remains (if i <n is satisfied), the processing from step 505 onward is repeated, and if the last character has been restored (i = n), the processing ends. In the example used here, as shown in FIG.
S [1] = 'a', S [2] = 'b', S [3] =
'r', S [4] = 'a', S [5] = 'c', S
[6] = 'a'. Therefore, the original input character string S = 'abraca' is restored.

As described above, in the test pattern compression apparatus 10 according to the present embodiment, when test pattern data to be compressed is input, BW is performed on all of the input data.
Rather than performing conversion, the number of transitions is less than a predetermined value,
Run-length encoding is directly performed on input data without performing BW conversion on input data for which data compression cannot be expected by BW conversion. Therefore,
It is possible to reduce the number of relatively time-consuming BW conversions, and to speed up data compression processing in the entire test pattern compression apparatus 10.

[Second Embodiment] In the test pattern compression apparatus 10 of the first embodiment described above, whether or not to perform BW conversion is determined based on the number of transitions of input data.
If the number of transitions of the input data becomes a large value to some extent, it is not known whether the number of transitions of data decreases or increases by the BW conversion alone, so that the BW conversion is performed based on information other than the number of transitions of the input data. It may be determined whether or not.

FIG. 12 is a diagram showing a configuration of a test pattern compression apparatus according to a second embodiment of the present invention. As shown in the drawing, the test pattern compression device 10A of the present embodiment includes a BW conversion application determination unit 20, a BW conversion unit 12,
Comparison unit 13, data selection unit 14, run-length compression unit 1
5 is included. The test pattern compression device 10A transmits the BW conversion application determining unit 11 shown in FIG.
The configuration has been replaced with a W conversion application determination unit 20.
The BW conversion application determining unit 20 performs an S conversion process on the input data and determines whether or not to perform the BW conversion based on the result, and determines the number of transition times measurement unit 21, the S conversion unit 22,
The number of transitions calculation unit 23 and the comparison unit 24 are included.

The S-transform is a fast transform algorithm derived from the BW transform and was published by Michael Schindler. For example, M. Schindler, “A Fast Block-sorting
Algorithm for Lossless Data Compression, ”Data Co
mpression Conference 1996, Snowbird Utah, March 1996
It is described in. This S-conversion improves the conversion processing speed by limiting the length of a character string (context) in sorting. The specific contents of the S conversion will be described later.

The transition count measuring section 21 measures the transition count φ _IN of the input data. The S conversion unit 22 converts the input data into S
Then, the number of transitions φ _S of the data after S conversion is measured.
The number-of-transitions calculator 23 calculates the number of transitions after BW conversion based on the number of transitions φ _S measured by the S converter 22. In general, the number of transitions φ _S of data after S conversion is
Since the length of the context is limited, the number of transitions after BW conversion φ
Becomes larger than _BW . Therefore, the number of transitions φ _S of data after S conversion can be obtained by multiplying the number of transitions φ _BW of data after BW conversion by a predetermined constant k (> 1). The constant k is the number of transitions φ _BW after BW conversion and the number of transitions φ after S conversion performed on many data.
It is preferably obtained from the result of comparison between _{S (φ S = k × φ} BW) statistically. For example, the average number of transitions after S conversion is B
For data that is twice the average number of transitions after W conversion, k = 2.

The comparison unit 24 compares the number of transitions φ _IN of the input data measured by the number-of-transitions measurement unit 21 with the number of transitions φ _S / k of the data after BW conversion calculated by the number-of-transitions calculation unit 23. Then, 1 or 0 is set to the flag f1 based on the comparison result. If φ _IN > φ _S / k, 1 is set to the flag f1, and φ _IN ≦ φ _S / k
In this case, 0 is set in the flag f1.

The components other than the BW conversion application determining unit 20 basically perform the same operations as the corresponding components of the first embodiment. That is, the BW conversion unit 12
When the flag f1 is set to 1 by the W conversion application determining unit 20, BW conversion is performed on the input data, and the number of transitions φ _BW of the data after the BW conversion is measured. Further, the comparing unit 13 compares the number of transitions φ _{IN of} the input data with the number of transitions φ _BW of the data after the BW conversion, and the data selector 14 compares the input data and the B
One of the data after W conversion is selectively output. Then, the run-length compression unit 15 performs run-length encoding on the data output from the data selection unit 14 to compress the data.

The above-described BW conversion unit 12 serves as a first sorting unit, the data selection unit 14 serves as a data selection unit, the BW conversion application determination unit 20 serves as a sorting operation application determination unit, and the S conversion unit 22 serves as a second sorting unit. The number-of-transitions calculation unit 23
Respectively correspond to the first transition number prediction means.

Next, the operation when the test pattern is compressed using the test pattern compression apparatus 10A of the present embodiment will be described. The test pattern compression device 10A performs various processes using data of a predetermined size L as one input unit, and data larger than L is divided into a plurality of sub-data of size L in advance, and the divided data is divided. Each sub data is input to the test pattern compression device 10A.

FIG. 13 is a flowchart showing an operation procedure of a test pattern compression operation performed in the test pattern compression apparatus 10A of the present embodiment. First, the transition count measuring unit 21 measures the transition count m of the input test pattern data and sets the value to φ _IN (step 60).
1). After performing the S conversion process on the input data (step 602), the S conversion unit 22 measures the number of transitions t of the S-converted data, and sets the value to φ _S (step 603). . In the S conversion process, the length of the context to be rearranged can be set arbitrarily.
May be a fixed value or may be input from the outside and set to an arbitrary value.

Next, the number-of-transitions calculator 23 calculates an estimated value (predicted value) of the number of transitions after the BW conversion using the number of transitions φ _S of the data after the S conversion and the constant k, and calculates the result φ _S / k is set to φ (step 604). Here, the constant k may be a fixed value, or may be input from the outside and set to an arbitrary value. Next, the comparison unit 24
Is the number of transitions φ _IN of the input data and the number of transitions calculator 23
Number of transitions of data after BW conversion estimated by
(= Φ _S / k), and the flag f1 is set based on the comparison result (step 605). Specifically, the comparison unit 24 sets 0 to the flag f1 when φ _IN ≦ 1 or φ _IN ≦ φ _S / k, and sets 1 to the flag f1 when φ _IN > φ _S / k.

Next, the BW conversion unit 12 determines whether or not the flag f1 is set to 1 (step 606). If the flag f1 is 1, the BW conversion processing is performed on the input data. After performing (step 607), the BW
The number of transitions n of the converted data is measured and the value is φ _BW
(Step 608). On the other hand, when the flag f1 is 0, the BW conversion unit 12 sets the input data size L, which is the upper limit value of the number of transitions, to φ _BW (step 609).

Next, the comparing section 13 compares the values of φ _IN and φ _BW and sets a flag f2 based on the comparison result (step 610). Specifically, the comparison unit 13 determines that φ _IN ≦ φ
_In the case of _BW , 0 is set to the flag f2, and in the case of φ _IN > φ _BW , 1 is set to the flag f2. Next, the data selection unit 14
It is determined whether the flag f2 is 1 (step 61).
1) If the flag f2 is 1, BW-converted data is selected (step 612); if the flag f2 is 0, input data is selected (step 613).

Next, the run-length compression unit 15 performs a compression process using run-length encoding on the data selected by the data selection unit 14 (step 61).
4) The compressed data is output (step 615), and a series of data compression processing ends. In order to decompress the compressed data to the original test pattern data, the compressed data output from the run-length compression unit 15 may include a check whether the run-length encoded compressed data is BW-converted. A flag f2 is added to indicate whether this is the case.

The data compressed by the test pattern compression apparatus 10A of the present embodiment in this manner is added to the compressed data similarly to the compressed data output from the test pattern compression apparatus 10 of the first embodiment. Flag f
When 2 is 1, the original test pattern data can be completely decompressed by performing BW inverse transformation after run-length decoding, and when the flag f2 is 0, only run-length decoding is performed. It can be completely expanded to the original test pattern data.

Next, the S conversion process performed at the time of data compression will be described based on a specific example. FIG.
9 is a flowchart illustrating an operation procedure of an S conversion process performed in the S conversion unit 22. First, the S conversion unit 22 acquires the input character string S, and sets the size (length) of the input character string to n (Step 701). In the following, for example, the description will be made using “abaabc” as the input character string S of 6 characters where n = 6.

Next, the S conversion unit 22 lists the n characters of the obtained input character string S and the context immediately after those characters (step 702). Here, the length of the context to be enumerated can be set arbitrarily. In the following, it is assumed that the length of the context is 2, for example. Further, the context refers to a character string arbitrarily and cyclically extracted from the input character string S. In the example of the character string S ('abaabc'), as shown in FIG. Six contexts are obtained.

Next, the S conversion unit 22 rearranges the six contexts obtained in step 702 in the order registered in the dictionary, and also rearranges the character immediately before the context together with the context (step 703). In the example of S = 'abaabc',
First, the context is rearranged such as 'aa' and then 'ab', and the character immediately before is rearranged as shown in FIG. At this time, if there are a plurality of the same contexts, the later contexts are arranged earlier in the input character string S. For example, in the example of the context 'ab' shown in FIG. 16, the second context 'ab' is the context immediately after the sixth character 'c' of the character string S, and the third context 'ab' is the character The context immediately after the third character 'a' in column S.

Next, the S conversion unit 22 determines the character immediately before the context (the immediately preceding character string P) rearranged in step 703,
The first two characters (equal to the length of the context) of the input character string S are obtained as the initial character string I (steps 704, 70).
5). In the example of S = 'abacabc', 'bcaab' is obtained as the immediately preceding character string P and 'ab' is obtained as the initial character string I, as shown in FIG. Thus, S =
The character string “abaabc” is the immediately preceding character string P = ′ b
aaaab 'and the initial character string I =' ab '. Finally, the S conversion unit 22 outputs the S-converted data (step 706), and ends a series of processes related to the S conversion.

As described above, in the test pattern compression apparatus 10A of the present embodiment, when test pattern data to be compressed is input, B
If the number of transitions of the data after the BW conversion estimated by performing the S conversion on the input data instead of performing the W conversion becomes larger than the number of transitions of the input data, B
A run-length encoding process is directly performed on input data without performing W conversion. Therefore, it is possible to reduce the number of relatively time-consuming BW conversions, and to speed up the data compression processing in the entire test pattern compression apparatus 10A.

By the way, S for the data of unit length L
Convert processing time to T_S, B for data of the same length L
The processing speed of W conversion is T_BWAnd Unit length of input data
L is divided into N sub data, and each divided sub data is
Data using BW transform and run-length coding
If compression processing is to be performed by
Conventional compression method combined with length coding
Then, N × (T_BW+ T_RLC) Time was needed. This
Where T_RLCIs the run length for the data of unit length L
Represents the processing time of Guss encoding. On the other hand, S transformation
In the compression method according to the present embodiment, N × T_S+ (N−n) × T_BW+ N × T_RLC  = N × (T_S+ T_BW+ T_RLC) −n × T_BW Time is needed. Here, n indicates that the BW conversion is being applied.
Indicates the number of stops. Therefore, the test pattern of the present embodiment is
By using the turn compression device 10A, nxT_BW−N × T_S Time can be reduced. Normally, S conversion is BW conversion.
BW conversion because it is tens to hundreds of times faster than conversion
(N × T)_BW) Only reduce processing time
It is possible to

The following is a comparison example of the processing time between the conventional compression method using only BW conversion and run-length coding and the compression method of the present embodiment using BW conversion, run-length coding and S conversion. In addition, in actual measurement, 100
A total of eight test pattern data of three test pattern data for the CICS pin microcomputer, two test pattern data for the 144 pin disk controller, and three test pattern data for the 144 pin RISC microcomputer were used.

FIG. 18 is a diagram showing a result of comparing the number of transitions of data after BW conversion with the number of transitions of data after S conversion using test pattern data used in an actual integrated circuit test as input data. . As shown in FIG.
The average number of transitions of the converted data becomes smaller as the context length L used for rearrangement becomes longer. When the context length L is 8, the average number of transitions of the data after the BW conversion becomes about 2.5 times.
When the context length L is 16, it is about 1.8 times the average number of transitions of the data after the BW conversion. A constant k for estimating the number of transitions of data after BW conversion from the number of transitions of data after S conversion can be set based on these scaling factors.

FIG. 19 is a diagram showing the results of comparing the conversion speeds of BW conversion and S conversion with test pattern data used in an actual integrated circuit test as input data. As shown in the figure, the speed of the S conversion is about 10 times that of the BW conversion depending on the length of the context (context length L = 16).
100 times (context length L = 8), 680 times (context length L =
1) Very fast. Therefore, according to the test pattern compression method of the present embodiment in which the number of transitions after the BW conversion is predicted using the S conversion, a significant reduction in the compression processing time can be expected as compared with the conventional compression method.

FIG. 20 is a diagram showing the result of comparing the processing time of the compression method of the present embodiment with the processing time of the conventional compression method combining BW conversion and run-length encoding. In the figure, K1, K2, and K3 are test patterns for the CISC microcomputer, A1 and A2 are test patterns for the disk controller, and S1, S2, and S3 are R.
It corresponds to each test pattern of the ISC microcomputer. Also, the measurement result of the present embodiment when the S-conversion is performed with the context length L set to 8 is indicated by hatching. As shown in FIG. 20, according to the test pattern compression method of the present embodiment using the S-transform, the processing time can be reduced by about 72% on average as compared with the conventional test pattern compression method.

[Third Embodiment] In the test pattern compression apparatus 10A of the second embodiment described above, the number of transitions of data after BW conversion is predicted based on the number of transitions of data after S conversion. Instead of measuring the number of transitions of the converted data, the entropy of the input data may be calculated, and the number of transitions of the data after the BW conversion may be predicted based on the calculation result.

FIG. 21 is a diagram showing a configuration of a test pattern compression apparatus according to a third embodiment to which the present invention is applied. As shown in the drawing, the test pattern compression device 10B of the present embodiment includes a BW conversion application determination unit 30, a BW conversion unit 12,
Comparison unit 13, data selection unit 14, run-length compression unit 1
5 is included. The test pattern compression device 10B transmits the BW conversion application determining unit 10 shown in FIG.
The configuration has been replaced with a W conversion application determination unit 30.
The BW conversion application determining unit 30 performs entropy calculation of the input data and determines whether or not to perform BW conversion based on the result, and determines the number of transition times measurement unit 31, the entropy measurement unit 32, and the number of transitions prediction unit 33. , And a comparing unit 34.

The transition count measuring section 31 measures the transition count φ _IN of the input data. The entropy measuring unit 32 measures the entropy H of the input data. Entropy H
Is the probability of occurrence of each symbol appearing in the input data, p _i (i =
1, ..., N) (N is the number of types of symbols appearing in the input data)

[0067]

(Equation 1)

Is calculated by

The number-of-transitions prediction section 33 calculates the entropy H calculated by the entropy measurement section 32,
The predicted value φ _{H of the} number of transitions after the BW conversion is calculated. The above-mentioned entropy H is obtained by averaging the probability of occurrence of each symbol in the input data for all symbols and adding a minus sign. Therefore, the fact that the value of the entropy H is small means that the same symbol is often included in the input data, and data having a small number of transitions can be obtained by performing BW conversion on the input data. Conversely, a large value of entropy H means that the same symbol is not included in the input data very much, and data having a large number of transitions can be obtained by performing BW conversion on the input data. As described above, there is a correlation between the entropy H and the number of transitions of the data after the BW conversion, and by using a predetermined prediction function F (H), the predicted value t (= F
(H)) can be calculated. Note that the prediction function F
Can be obtained from statistical analysis of entropy and number of transitions.

The comparison unit 34 compares the number of transitions φ _IN of the input data measured by the number-of-transitions measurement unit 31 with the predicted value φ _{H of the} number of transitions of the data after BW conversion obtained by the number-of-transitions prediction unit 33. Then, 1 or 0 is set to the flag f1 based on the comparison result. If φ _IN > φ _H , 1 is set to the flag f1, and if φ _IN ≦ φ _H , 0 is set to the flag f1.

The components other than the BW conversion application deciding unit 30 perform basically the same operations as the corresponding components of the first embodiment. That is, the BW conversion unit 12
When the flag f1 is set to 1 by the W conversion application determining unit 30, BW conversion is performed on the input data, and the number of transitions φ _BW of the data after the BW conversion is measured. Further, the comparing unit 13 compares the number of transitions φ _{IN of} the input data with the number of transitions φ _BW of the data after the BW conversion, and the data selector 14 compares the input data and the B
One of the data after W conversion is selectively output. Then, the run-length compression unit 15 performs run-length encoding on the data output from the data selection unit 14 to compress the data.

The BW conversion unit 12 described above is used as the first rearrangement unit, the data selection unit 14 is used as the data selection unit, the BW conversion application determination unit 30 is used as the rearrangement operation application determination unit, and the entropy measurement unit 32 is used as the entropy measurement unit. The transition number prediction unit 33 corresponds to the second transition number prediction unit.

Next, the operation of compressing a test pattern using the test pattern compression device 10B of the present embodiment will be described. The test pattern compression apparatus 10B performs various processes using data of a predetermined size L as one input unit, and data larger than L is divided into a plurality of sub-data of size L in advance, and the divided data is divided. Each sub data is input to the test pattern compression device 10B.

FIG. 22 is a flowchart showing an operation procedure of a test pattern compressing operation performed in the test pattern compressing apparatus 10B of the present embodiment. First, the transition count measuring unit 31 measures the transition count m of the input test pattern data and sets the value to φ _IN (step 80).
1). Further, the entropy measuring unit 32 measures the entropy h and sets the value to H (step 80).
2).

Next, based on the entropy H calculated by the entropy measuring section 32, the transition number prediction section 33 predicts the transition number prediction value t (t = t =
F (H)) is calculated and its value is set to φ _H (step 803). Next, the comparison unit 34 compares the number of transitions φ _IN of the input data with the number of transitions φ _H predicted by the number-of-transitions prediction unit 33, and based on the comparison result, sets the flag f1
Is set (step 804). Specifically, the comparison unit 34 sets the flag f1 when φ _IN ≦ 1 or φ _IN ≦ φ _H.
Is set to 0, and when φ _IN > φ _H , 1 is set to the flag f1.

Next, the BW conversion unit 12 determines whether or not the flag f1 is set to 1 (step 805). If the flag f1 is 1, the BW conversion processing is performed on the input data. After performing (step 806), the BW
The number of transitions n of the converted data is measured and the value is φ _BW
(Step 807). On the other hand, if the flag f1 is 0, the BW conversion unit 12 sets the input data size L, which is the upper limit value of the number of transitions, to φ _BW (step 808).

Next, the comparing section 13 compares the values of φ _IN and φ _BW and sets a flag f2 based on the comparison result (step 809). Specifically, the comparison unit 13 determines that φ _IN ≦ φ
_In the case of _BW , 0 is set to the flag f2, and in the case of φ _IN > φ _BW , 1 is set to the flag f2. Next, the data selection unit 14
It is determined whether the flag f2 is 1 (step 81).
0), if the flag f2 is 1, BW-converted data is selected (step 811), and if the flag f2 is 0, input data is selected (step 812).

Next, the run-length compression section 15 performs a compression process using run-length coding on the data selected by the data selection section 14 (step 81).
3) The compressed data is output (step 814), and a series of data compression processing ends. In order to decompress the compressed data to the original test pattern data, the compressed data output from the run-length compression unit 15 may include a check whether the run-length encoded compressed data is BW-converted. A flag f2 is added to indicate whether this is the case.

The data compressed by the test pattern compression apparatus 10B of the present embodiment in this manner is added to the compressed data similarly to the compressed data output from the test pattern compression apparatus 10 of the first embodiment. When the flag f2 is 1, the original test pattern data can be completely decompressed by performing the BW inverse transform after the run-length decoding, and when the flag f2 is 0, only the run-length decoding is performed. Thus, the original test pattern data can be completely expanded.

As described above, in the test pattern compression apparatus 10B of the present embodiment, when the test pattern data to be compressed is input, the B
Instead of performing W conversion, the entropy H of the input data
If the number of transitions of the data after the BW conversion predicted by calculating the number of times becomes larger than the number of transitions of the input data, the run-length encoding process is directly performed on the input data without performing the BW conversion. . Therefore, it is possible to reduce the number of relatively time-consuming BW conversions, and to speed up data compression processing in the entire test pattern compression apparatus 10B.

In each of the above-described embodiments, the case where the test pattern data for the integrated circuit is compressed has been described. However, the present invention can be applied to the case where data used for other purposes is compressed. .

[0082]

As described above, according to the present invention, it is determined whether or not the number of data transitions decreases when a rearrangement process for collecting symbols of the same or similar context included in input data is performed. It is possible to reduce the number of input data to be subjected to the rearrangement process by determining in advance without actually performing the rearrangement process and stopping the rearrangement process when it is determined that the rearrangement does not decrease. The speed of the compression process can be increased.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration of a test pattern compression device according to a first embodiment.

FIG. 2 is a flowchart showing an operation procedure of a test pattern compression operation performed in the test pattern compression apparatus shown in FIG.

FIG. 3 is a flowchart illustrating an operation procedure of a BW conversion process performed in a BW conversion unit.

FIG. 4 is a diagram showing a specific example of BW conversion.

FIG. 5 is a diagram showing a specific example of BW conversion.

FIG. 6 is a diagram showing a specific example of BW conversion.

FIG. 7 is a flowchart showing an operation procedure of a run-length encoding process performed in a run-length compression unit.

FIG. 8 is a flowchart showing an operation procedure of a run-length decoding process.

FIG. 9 is a flowchart showing an operation procedure of an inverse BW conversion process.

FIG. 10 is a diagram illustrating a specific example of inverse BW conversion.

FIG. 11 is a diagram illustrating a specific example of inverse BW conversion.

FIG. 12 is a diagram illustrating a configuration of a test pattern compression device according to a second embodiment.

13 is a flowchart showing an operation procedure of a test pattern compression operation performed in the test pattern compression device shown in FIG.

FIG. 14 is a flowchart showing an operation procedure of an S conversion process performed in the S conversion unit.

FIG. 15 is a diagram showing a specific example of S conversion.

FIG. 16 is a diagram showing a specific example of S conversion.

FIG. 17 is a diagram illustrating a specific example of S conversion.

FIG. 18 is a diagram illustrating a result of comparing the number of transitions of data after BW conversion and the number of transitions of data after S conversion.

FIG. 19 is a diagram showing the result of comparing the conversion speeds of BW conversion and S conversion.

FIG. 20 shows the processing time and BW of the compression method according to the second embodiment.
FIG. 11 is a diagram illustrating a result of comparing a processing time of a conventional compression method combining transform and run-length encoding.

FIG. 21 is a diagram illustrating a configuration of a test pattern compression device according to a third embodiment.

22 is a flowchart showing an operation procedure of a test pattern compression operation performed in the test pattern compression device shown in FIG. 21.

[Explanation of symbols]

 10, 10A, 10B Test pattern compression device 11, 20, 30 BW conversion application determination unit 12 BW conversion unit 13, 24, 34 Comparison unit 14 Data selection unit 15 Run-length compression unit 21, 31, Transition number measurement unit 22 S conversion unit 23 transition number calculation unit 32 entropy measurement unit 33 transition number prediction unit

Claims (14)

[Claims] A first rearranging unit for performing a rearranging operation for collecting symbols of the same or similar context included in input data; and a rearranging operation for the input data by the first rearranging unit. Is performed without performing the rearrangement operation to determine whether the number of data transitions is reduced, and when the reduction is determined, the first rearrangement unit performs the rearrangement operation. A sorting operation application determining unit that determines that the sorting operation is stopped by the first sorting unit when it is determined that the sorting operation does not decrease; and a sorting operation by the first sorting unit is performed. If not, the corresponding input data is output, and if the sorting operation by the first sorting means is performed, the corresponding input data and the first sorting data are output. Data selecting means for selectively outputting the data having a smaller number of transitions from the data after the data is rearranged, and a run length for performing a run length encoding process on the data output from the data selecting means. A data compression device comprising: encoding means. 2. The data compression device according to claim 1, wherein the first rearrangement unit performs rearrangement of data by Burrows-Wheeler transform processing. 3. The data compression device according to claim 1, wherein the input data is test pattern data for an integrated circuit test. 4. The rearrangement operation application determining unit according to claim 1, wherein the rearrangement operation application determination unit determines whether or not the number of data transitions is reduced by performing the rearrangement operation. A data compression device, wherein the data compression is performed based on the number of transitions. 5. The sorting operation application determining unit according to claim 1, wherein the rearrangement operation application determining unit collects symbols of the same or similar context included in the input data while limiting the length of the context. A second rearranging unit for performing a rearranging operation; and a first rearranging unit for the input data based on the number of transitions of data after the rearranging operation is performed by the second rearranging unit. A first transition number predicting unit for predicting the number of transitions of the data after the rearrangement operation is performed, based on the number of transitions predicted by the first transition number predicting unit. A data compression apparatus for determining whether or not the number of data transitions is reduced by performing a rearrangement operation by a rearranging unit. 6. The data compression device according to claim 5, wherein the second rearrangement unit rearranges the data by an S conversion process. 7. The reordering operation application determining unit according to claim 1, wherein the rearrangement operation application determining unit determines entropy of the input data based on an entropy value measured by the entropy measuring unit. And the first
A second transition number prediction unit that predicts the number of transitions of the data after the rearrangement operation is performed by the rearrangement unit, and based on the transition number predicted by the second transition number prediction unit. A data compression apparatus for determining whether or not the number of data transitions is reduced by performing a rearrangement operation by a first rearrangement unit. 8. A data compression method that combines a first rearrangement process for collecting symbols of the same or similar context included in data and a run-length encoding process, wherein the first rearrangement process is performed. A first step of determining whether or not the number of data transitions is reduced by performing the first rearrangement process; and a determination is made in the first step that the number of transitions is reduced. A second step of actually performing the first rearrangement processing on the input data; and a step of performing the first rearrangement processing when the first rearrangement processing is performed in the second step. Comparing the number of transitions of the data obtained by the replacement process with the number of transitions of the input data, and selectively outputting data having a small number of transitions, and Outputting the input data when it is determined that the number of transitions does not decrease;
And a fourth step of performing the run-length encoding process on the data output in the third step. A data compression method, comprising: 9. The data compression method according to claim 8, wherein the first rearrangement process performed in the second step is a Burrows-Wheeler transform process. 10. The data compression method according to claim 8, wherein the input data is test pattern data for an integrated circuit test. 11. The data compression method according to claim 8, wherein the determining operation in the first step is performed based on the number of transitions of the input data. 12. The fifth rearrangement process according to claim 8, wherein a second rearrangement process for collecting symbols of the same or similar context included in the input data while limiting the length of the context is performed. And the number of data transitions obtained by the second rearrangement process performed in the fifth step,
A sixth step of predicting the number of transitions of the data obtained by the first rearrangement processing. The method according to the first step, further comprising: A data compression method characterized by performing a determination operation. 13. The data compression method according to claim 12, wherein the second rearrangement process performed in the fifth step is an S conversion process. 14. The method according to claim 8, wherein a seventh step of measuring entropy of the input data, and the first step is performed based on a value of the entropy measured in the seventh step. An eighth step of predicting the number of transitions of the data obtained by the rearrangement process, and performing the determination operation in the first step based on the number of transitions predicted in the eighth step. A data compression method, characterized in that: