JPH10341163A - Data compression method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To allocate the least necessary number of bits to the characters and also to decrease the quantity of data by inputting a character string where a single character can have four states at most and converting each character of the character string into a quadruple notation to generate and output a new character string. SOLUTION: The meanings of characters of a source list 1 are shown as character '1' → logic 1, character '0' → logic 0, character 'L' → an L level, character 'H' → an H level, character 'P' → a positive polarity pulse, character 'N' → a negative polarity pulse and character 'X' → the indefiniteness of state respectively together with each singularity. Then a conversion table 2 is produced in consideration of the said singularity. When the 1st line, for example, of the list 1 is converted on the table 2, a character string '11123003' is acquired in a field (b) of the table 2 owing to 1 (logic) → 1, P (positive polarity pulse) → 2, N (negative polarity pulse) → 3, L (L level) → 0 and X (indefinite state) → 3 respectively.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a data compression method, and more particularly to a data compression method suitable for use in a test pattern of a semiconductor integrated circuit. In general, test patterns that are treated as binary data in the execution state are ASCII (Am
erican Standard Code for Information Interchange)
And EBCDIC (Extended Binary Coded Decimal Inte
rchange Code). For example, each one-byte readable character string "ABCDEF" is represented by A
"41, 42, 43, 4" in hexadecimal notation of the SCII code
4,45,46 ", and" p100000 in binary notation
1, p1000010, p1000011, p1000
100, p1000101, p1000110 "(p is a parity bit).

[0002] The data amount of the character string "ABCDEF" is given by the total number of bits in a binary representation. Specifically, the number of bits of the binary code × the number of characters in the character string = 8 × 6 = 48 bits. However, an extremely large number of characters such as a test pattern of a semiconductor integrated circuit results in an extremely large amount of data. In some cases, data storage, transfer, or processing may be hindered, and there is a need for a technology that can efficiently compress source test patterns to reduce the amount of data.

[0003]

2. Description of the Related Art FIG. 6 is a diagram for explaining a conventional data compression method. In this figure, symbols such as △, Δ, and × are arbitrary characters, and here, it is assumed that a data sequence of one line in the list is composed of seven characters for convenience. "REPn" and "END" in the list are
REPn is a control character for data compression, and REPn is a data sequence from the next line to the END tag (in the example shown, 図 示 ○
Xx) n times. According to this, a higher compression ratio can be obtained as rows with the same character string (redundant rows) appear successively.

[0004]

However, while such a conventional data compression method is useful in that "redundant rows" can be compressed, its compression effect is determined solely by the frequency of occurrence of redundant rows. There is a problem that almost no compression effect can be obtained for a source list with few redundant lines.

Accordingly, an object of the present invention is to obtain a good compression effect without depending on the frequency of appearance of redundant rows.

[0006]

The invention according to claim 1 is
A character string in which one character can take up to four states is input, and each character in the character string is converted to a quaternary notation to generate and output a new character string. According to this, the minimum necessary number of bits can be assigned to a character that never takes more than four states, and the data amount can be reduced.

According to a second aspect of the present invention, in the first aspect of the present invention, the new character string is converted into hexadecimal notation, and the same character that appears continuously in the converted character string is replaced by one character. And expressing the number of the same characters using unused characters in hexadecimal notation. According to this, the redundant characters in the character string can be replaced with unused characters (G to) in hexadecimal notation, and the data amount can be further reduced particularly when the number of characters is large.

[0008]

1 to 5 show an embodiment of a data compression method according to the present invention. First, the specificity of the test pattern of the semiconductor integrated circuit will be described. A test pattern of a semiconductor integrated circuit includes an input signal sequence to a circuit to be tested and an expected value signal sequence (hereinafter, an output signal sequence) to be output in response to the input signal sequence.

The states of each signal in the input signal series are basically two states of logic 0 and logic 1, and even if the polarity state of the signal (positive or negative) is added to this state, only four states at maximum. Not possible. The same applies to the state of each signal in the output signal series, and only four states of an L-level output state, an H-level output state, a high-impedance output state, and an undefined state can be taken.

The source test pattern is data composed of a combination of the above 4 + 4 (maximum) states. However, since debugging after data generation is indispensable, it is coded with, for example, readable characters as shown in FIG. FIG.
In the source list 1 of this example, one character string is composed of eight characters, but this is a convenience value. The actual number of characters is determined by the number of pins of the circuit under test.

In FIG. 1, the meaning of each character is
It is as follows, character "1" → logical 1 character "0" → logical 0 character "L" → L level character "H" → H level character "P" → positive polarity pulse Character "N" → negative polarity pulse Character “X” → undefined state From the description of the specificity, for example, it can be directly read that the character “1” is an input signal sequence and the character “L” is an output signal sequence.

The present embodiment is provided with the following characteristic procedure to reduce the amount of data for each line of the source list 1. (1) Procedure 1 First, the first line of the source list 1 of FIG. 1 is read, and converted into numerical values in order from the first character using the conversion table 2 of FIG. The conversion table 2 in FIG. 2 is created in consideration of the above-described specificity. For example, when the field a is referred to in each state of the input / output signal sequence, the numerical values 0 to 3 corresponding to the state are obtained. It is taken out from the field b. When the first row of the source list 1 in FIG. 1 is converted using the conversion table 2, 1 (logic 1) → 1 P (positive pulse) → 2 N (negative pulse) → 3 L (L level) → Since 0 X (undefined state) → 3, “11” is obtained from the field b of the conversion table 2.
A character string of “123003” is obtained. However, input / output cannot be distinguished by using this character string alone. For example, “N” and “X” before conversion cannot be distinguished by looking at 3 after conversion. ..., a new attribute character string is generated.
The field c of the conversion table 2 is for generating an attribute character string. Using this information, "00000
An attribute character string of “111” is obtained.Five “0” represent an input attribute, and three “1” represent an output attribute.
That is, five “0” s match the input attribute of the first line of the source list 1 (“111PN”), and three “1” s match the output attribute of the first line of the source list 1 (“LLX”). Therefore, one character string of the source list 1 can be expressed by a pair of the conversion character string and the attribute character string. FIG. 3 is a conversion result of the source list 1 of FIG. 1 (hereinafter, a first intermediate code list 3). (2) Procedure 2 By the way, each digit of the first intermediate code can take only numerical values of 0, 1, 2, and 3. That is, since a numerical value of "4" or more never appears, it is useless to apply a 1-byte (8-bit) code (for example, an ASCII code) to such a sequence.

In the present embodiment, attention is paid to the fact that each numerical value of the first intermediate code is a quaternary number which can only take the values of 0, 1, 2, and 3. The amount of data is reduced by converting the data into the described second intermediate code. For example, the first intermediate code “11123”
“003” is “0101 01 10” in binary quaternary notation.
11 00 00 11 ", which suffices for 16 bits. That is, in the case of a 1-byte code, the data amount is 48 bits, so that the data amount is 1/3.

If the above-mentioned binary coded quaternary number is expressed in hexadecimal notation, it becomes "56C3". Similarly, if other character strings are also converted to binary coded quaternary notation, and converted to hexadecimal notation, as shown in FIG. A second intermediate code list 4 is obtained. FIG.
6C3 "," 0015 "," 42D7 "and" 001 ".
The four hexadecimal codes of "5" are shown. "56C3" and "0015" on the first line correspond to "111PNLLX" of the source list 1 (see FIG. 1), and "42PLLX" on the second line.
D7 ”and“ 0015 ”are“ 100 ”of the source list 1
PNHHX ".

The total number of bits of the second intermediate code list 4 is:
It is given by the value obtained by multiplying the number of hexadecimal codes (16 bits) by the number of bits (16 bits) of one hexadecimal code. That is, in the example of FIG. 4, 16 bits × 4 = 64 bits. On the other hand, the total number of bits of the original source list 1 (FIG. 1) is, as described at the beginning, the number of bits of the binary code × the number of characters of the character string = 8 × 16 = 128 bits. Of the second intermediate code is 64/1
28 = 1/2, which is halved. (3) Procedure 3 Although practically sufficient data compression effect can be obtained by only the above procedure 1 and procedure 2, it is desirable to execute the following procedure to obtain a further compression effect.

That is, the second intermediate code in FIG. 4 is in hexadecimal notation, but the hexadecimal code is a numerical value from 0 to 9 and A to
Only letters up to F are used. Any further letters (G
~) Are unused. In step 3, the unused codes (G-) are used to compress the redundant code in one line. The compression effect obtained by the procedure 3 increases as the number of codes per line increases, in other words, as the number of pins of the circuit under test increases. This is because the redundancy increases accordingly.

[0017] Now, one of "00000115555"
Assume a one-digit hexadecimal code. In this example, one of the five
Hexadecimal code "0" continues, two hexadecimal codes "1"
And four hexadecimal codes “5” are continuous.
That is, conversion into the form of “(5) 0 (2) 1 (4) 5”... Where the numerical value in parentheses is a redundant number shown for convenience.
If possible, further data reduction can be expected. However, if this is simply replaced by a sequence, "502145"
And it is not possible to distinguish between the redundant number and the actual number. Therefore, in the procedure 3, the redundancy table 5 as shown in FIG. 5 is prepared, and the redundancy number in the parentheses is replaced by using this table. In the case of the above example, the sequence after replacement is “K0H
1J5 ", which makes it possible to distinguish between the redundant number (K, H, J) and the actual number (0, 1, 5). In addition, since the number of characters decreases from 11 to 6, even if 1-byte characters are used, , 8 bits × 6 characters = 48 bits, which is about 1 bit smaller than 16 bits × 11 = 176 bits before redundant replacement.
/3.6 data compression effect can be obtained. In the case of a redundant number of 2 (in the above example, two consecutive hexadecimal codes "1") are not affected by compression even if they are replaced, so that (H1 → 11) is used as it is to avoid unnecessary processing. It does not matter.

[0018]

According to the present invention, the minimum necessary number of bits can be allocated to a character that does not take more than four states, and the amount of data can be reduced. In addition, the redundant characters in the character string can be replaced with unused characters (G to) in hexadecimal notation, and the data amount can be further reduced particularly when the number of characters is large.

[Brief description of the drawings]

FIG. 1 is a source list diagram of one embodiment.

FIG. 2 is a conversion table diagram of one embodiment.

FIG. 3 is a diagram illustrating a first intermediate code list according to one embodiment;

FIG. 4 is a diagram illustrating a second intermediate code list according to an embodiment;

FIG. 5 is a diagram illustrating a redundancy table according to an embodiment;

FIG. 6 is a conventional source list diagram.

[Explanation of symbols]

 1: Source list 2: Conversion table 3: First intermediate code list 4: Second intermediate code list 5: Redundancy table

Claims (2)

[Claims] 1. A character string in which one character can take up to four states is input, and each character in the character string is converted to a quaternary notation to generate and output a new character string. Data compression method. 2. Converting the new character string into hexadecimal notation, leaving only one identical character appearing continuously in the converted character string, and removing unused characters in hexadecimal notation. 2. The data compression method according to claim 1, wherein the number of the same characters is expressed by using the same.