JPS62216033A - Data conversion processing system - Google Patents

Abstract

PURPOSE:To decrease the number of function parts for conversion of data by using as intervention the floating decimal point data containing the exponent value having 2<8> as its base when the conversion is carried out between several types of floating decimal point data and the pack decimal data. CONSTITUTION:The floating decimal point data 1 and 2 containing the exponent value having 2 as its base or the decimal point data 3 and 4 containing the exponent value having 2<4> as its base are converted into the pack decimal data 5. In such a case, the mantissa data of those data 1-4 are shifted so that 2<8> is defined as a base and then the exponent value is changed to secure a base of 2<8>. Thus the intermediate data 6 serving as the floating decimal point data containing the exponent value having 2<8> as its base is obtained and then multiplied by '10' for an integer part of the mantissa part. While the decimal part of the mantissa data is divided by '10' to obtain the pack decimal data 5. In the same way, the data 5 is converted into the data 6 when the data 5 is converted into the data 1-4. Thus the desired floating decimal point data is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] A data processing device that performs data conversion between any two of floating point data of a first group, floating point data of a second group, and packed decimal data.

The number of functional units for data conversion is provided with a conversion processing unit that obtains intermediate data that is floating point data having an exponent value with base 2B, and is configured to perform the data conversion using the intermediate data, and the number of functional units for data conversion. A data conversion processing method is disclosed that reduces the amount of data.

[Industrial application field]

The present invention relates to a data conversion processing method, particularly a data conversion processing that involves a process of obtaining intermediate data when converting data between any two types of floating point data or packed decimal data. Regarding the method.

[Conventional technology]

Conventionally, in a data processing device, as shown in FIG. "Single precision floating point data with an exponent value of base 2'" 3. "Double precision floating point data with an exponent value of base 24" 4. "Back decimal data" 5 can be used respectively. There is mutual data conversion as indicated by the arrows in the figure, as necessary.

In the case shown above, data 1 and data 2 are referred to as the first group of floating point data in this specification, and data 3 is referred to as the first group of floating point data.
and data 4 are called the second group of floating point data. The number of bits of the exponent part and the number of bits of the mantissa part in each of these data are summarized in FIG. 7, respectively.

As can be seen from FIG. 7, data 1 and data 2 have different numbers of bits in their exponent parts, but data 3 and data 4 have the same number of bits in their exponent parts. In consideration of this, in the past, a total of seven conversion function units were prepared as indicated by arrows in FIG. 6 for performing conversion.

Figure 8 shows the data format of packed decimal data and floating point data (in the case shown, data 2 is an example).
4 bits of 9 bytes constitute the code part.

It also has 2 bytes of binary data indicating the decimal point position. In the case of the floating point data shown in Figure 8(B), 1 bit of the #7 byte constitutes the sign part, the following 11 bits constitute the exponent part, and the following 52 bits constitute the mantissa. It makes up the department. When the content of the illustrated sign part is S, the content of the exponent part is e, and the content of the mantissa part is f, the floating point data is (1)3 *2@-10! It can be expressed as ff*(1, f).

[Problem that the invention seeks to solve]

Conventionally, as explained in connection with FIG. 6, it is necessary to perform conversion between each data, and it is necessary to prepare a data conversion function unit for performing each conversion. If data in a format other than that shown in FIG. 6 is used, it is necessary to prepare a data conversion function unit between the data and each data.

[Means for solving problems]

The present invention solves the above points, and provides a process for obtaining floating point data with an exponent value of base 28 when data conversion between various types of floating point data and packed decimal data is performed. Focusing on the existence of , floating-point data with an exponent value with the base of 28 is defined as intermediate data, and a conversion function unit between various types of the above data and the intermediate data is prepared. .

FIG. 1 shows a basic configuration diagram of the present invention. The code l in the figure is “
"Single-precision floating point data with an exponent value to the base 2",
2 is "double-precision floating-point data with an exponent value of base 2", 3 is "single-precision floating-point data with an exponent value of base 24", and 4 is "with an exponent value of base 24". 5 is packed decimal data, and 6 is intermediate data, which represents "floating point data with an exponent value having a base of 2°." Also, the code ff to 11
each represents a mutual conversion function section. Note that the intermediate data 6 has 2 bytes (16 bits) as an exponent part and 8 bytes (64 bits) as a mantissa part.

[Effect]

In converting data 1, data 2, and data 4 shown in Figure 1 into intermediate data 6, (i) Data 1 and data 2
Shift the mantissa data of data 4 so that the base is 28, and (ii) change the exponent value so that the base is 28 (
(divide by 8) and (iii) obtain intermediate data 6. Then, regarding the intermediate data 6 obtained in this manner, the integer part of the mantissa data is multiplied by 10, and the decimal part is divided by 10 to obtain packed decimal data 5.

In addition, when converting packed decimal data 5 to intermediate data 6, (i) convert packed decimal data to binary data by treating it as an integer value including data below the decimal number, and (
ii) Shift the data from low to high until the most significant byte is no longer O, then subtract the exponent value by the shifted number, (iii) Divide by 10 for the number of decimal point positions (convert to correct binary data) ). Then, the desired floating point data is obtained. That is, floating point data is created by combining exponent data and binary data.

〔Example〕

FIG. 2 shows a processing flow for converting punctured decimal data into floating point data, and FIG. 3 shows a processing flow for obtaining banked decimal data from floating point data. That is, FIG. 2 corresponds to the processing flow of the conversion function unit 11→(8 or 9 or 10) shown in FIG. Further, FIG. 3 corresponds to the processing flow of the conversion function unit (8 or 9 or 10)→11 shown in FIG.

4(A) and 4(B) are explanatory diagrams specifically explaining the flow shown in FIG. 2. FIG. The figure represents a situation in which packed decimal data with a value r12.5J is converted to double precision floating point data with a base 2 exponent value.

Reference numeral 20 in the figure represents packed decimal data with a value of 125, and 21 represents binary data indicating the decimal point position. Reference numerals 22 and 23 represent mantissa data and exponent data obtained by converting data 20, including data below the decimal number, into binary data, which is regarded as an integer value r125J. Reference numerals 24 and 25 represent the result of shifting the mantissa data 22 in units of bytes until the most significant byte is no longer O, and the result of changing the exponent value of the most significant byte corresponding to the shift. Further, 26, 27° and 28 form a set and represent intermediate data 6.

Note that data 26 is obtained by dividing data 24 by 10 by the number of times at the decimal point indicated by data 21. Further, data 27 is index data calculated from the above data 25, and data 28 represents code data.

Reference numeral 29 represents the binary representation of the data 26. The code 30 is the data 29 for converting to floating point data.
It shows the result of shifting .

Reference numeral 31 is index data, and the index value of intermediate data (
Multiply the value of data 27) by 8 and add 7 to it.
+7), the amount shifted in the conversion process from data 29 to data 30 is subtracted (7-4.), and the so-called geta (value 1023) is removed (3+1023). and.

Reference numeral 32 is floating point data obtained by combining data 30 and data 31. Note that the code 33 represents the data 32 as 1
It is expressed in hexadecimal notation.

FIG. 5 is an explanatory diagram specifically explaining the flow shown in FIG. 3. The figure represents the situation in which packed decimal data 5 is obtained from data 2 with the value r12.5J.

Reference numerals 32 and 33 in the figure represent those corresponding to FIG. Reference numerals 34, 35, and 36 indicate data 32 decomposed into a sign part, an exponent part, and a mantissa part. Reference numeral 37 is the result of removing geta (value 1023) from data 35, and indicates an exponent value for the omitted bit.

The code 38 is the result of converting the most significant byte into an exponent value, and the base is set to 28 at this time. Reference numeral 39 represents the result of shifting the data to 5 data 36 so that 2 n becomes the base. Note that illustrated data 38 and 39 correspond to intermediate data.

The code 40 is the integer part extracted from the data 39, and the integer part data is divided by 10.

This is integer data extracted in external decimal form. That is, in the case shown in 5.

00001100 ÷ 00001010 is performed to obtain a quotient of 00000001 and a remainder of 00000010, and the remainder is converted to a decimal number to obtain the value "F2". Further, 00000001 ÷ 00001010 is performed to obtain the quotient oooooooo and the remainder 00000001, and the remainder is converted to a decimal number to obtain the value "Fl".

The code 41 is the decimal part of the data 39.
In the case shown in 2, it is taken out while being multiplied by 0.

Do 10000000 x 00001010,
Result 000001010000... Obtain the 1-byte overflow part, convert the overflow part to a decimal number, and get the value rF
I'm getting 5J.

Reference numeral 42 represents the number of digits of the decimal point. and code 4
3 represents packed decimal data as a result of conversion.

〔Effect of the invention〕

As explained above, according to the present invention, it is possible to reduce the number of conversion function sections from seven to five, for example in the examples shown in FIGS. 1 and 6.

[Brief explanation of drawings]

Figure 1 is a diagram of the principle configuration of the present invention, Figure 2 is an example flow of converting packed decimal data to floating point data, and Figure 3 is an example process of converting floating point data to punctured decimal data. Flow, Figures 4 (A) (B) and 5
The figure shows an example of processing, FIG. 6 is an explanatory diagram for explaining the conventional case, and FIGS. 7 and 8 are explanatory diagrams for explaining type B of data representation. In the figure, 1.2 is the first group of floating point data, 3°4 is the second group of floating point data, 5 is the packed decimal data, and 6 is the floating point data of the second group.
denotes intermediate data, and F to 11 denote conversion function units, respectively. Patent applicant Fujitsu Ltd. Representative Patent Attorney Mori 1) Hiroshi (1 other person) Omi 6 each! ]
Moon's irises 1 and 71 12 town Kunikata 1 Endokoro, tl flow (III processing) 11) - (11η 3 Sengaku 2 Sengaku 舊!,'e Tsukizu ffl 171Xl

Claims (1)

[Claims] At least a first group of floating point data (1, 2) with an exponent value of base 2, and a second group of floating point data (1, 2) with an exponent value of base 2^4. 3, 4) and pack 10
A data processing device that performs data conversion to and from decimal data (5), comprising a conversion processing unit that obtains intermediate data (6) that is floating point data with an exponent value having a base of 2^8, and converts the intermediate data into The conversion processing unit that obtains the above includes a mutual conversion function unit (8, 9) between the intermediate data and the floating point data of the first group, and a mutual conversion function unit (8, 9) between the intermediate data and the floating point data of the second group. a mutual conversion function unit (10); and a mutual conversion function unit (11) between the intermediate data and the packed decimal data; A data conversion processing method characterized in that when converting data between any two of the decimal point data and the packed decimal data, the conversion is performed by intervening intermediate data from the conversion processing section. .