Packed decimal arithmetic



Yüklə 37,86 Kb.
tarix08.10.2017
ölçüsü37,86 Kb.
#3524


Packed decimal is a convenient format for doing many arithmetic calculations in assembly language for several reasons:
1) All computations occur in integer arithmetic (no decimals, 5/2 = 2, etc.),
2) Packed decimal fields are easy to read in a storage dump,
3) Computations occur in base 10.
The main disadvantage to packed decimal arithmetic is that decimal points are not stored internally. This means a programmer must keep up with decimals and make sure they are printed in the correct positions on any report.
In this topic we discuss commonly used packed decimal computations with examples. Later we will consider arithmetic operations with decimal points.
Copying Packed Decimal Fields
When copying a packed decimal field, be sure to use the Zero and Add Packed instruction, ZAP. By using ZAP, you are assured that the target field will be properly initialized. Many beginners make the mistake of using MVC when copying packed decimal values. This can lead to an error which is illustrated in the following example.
MVC AFIELD,PKFIELD AFIELD = X’038CFF’

ZAP AFIELD,PKFIELD AFIELD = X’00038C’

...

AFIELD DS PL3



PKFIELD DC PL2’38’ PKFIELD = X’038C’

DC X’FF’
We are copying a 2-byte packed field, PKFIELD, to a 3-byte field, AFIELD. Since MVC has an SS1 format, the length of AFIELD is used to determine that 3 bytes will be “moved” by this instruction. The effect of this instruction is to copy the 2 bytes in PKFIELD and the byte which follows PKFIELD as well. As a result, AFIELD does not contain a packed value. On the other hand, the ZAP above first initializes AFIELD with a packed decimal zero just before adding the packed value of PKFIELD. This produces the correct packed decimal value X’00038C’. This type of error with the MVC instruction occurs each time the fields involved have different sizes.


Care must be taken even when using a ZAP to copy a packed field. If the target field is too small to hold the result, high order truncation of digits can occur, causing an overflow. Consider the following example involving AFIELD defined above.
ZAP AFIELD,=P’123456789’ AFIELD = X’56789C’
After executing the instruction above, the high-order digits of the packed decimal literal have been truncated. This may or may not cause the program to abend, depending on the decimal-overflow mask. (See SPM.) In the case where the program continues execution, the programmer is not immediately aware that an error has occurred.
One side effect of executing a ZAP is that the condition code is set to indicate how the target field compares to zero. The condition code can be tested with the branch on condition instruction using the extended mnemonics. Here is an example,
ZAP FIELD1,FIELD1 SET THE CONDITION CODE

BZ WASZERO BRANCH IF ZERO



Adding Packed Decimal Fields
Next we consider the problem of adding several packed decimal fields. In doing this we must estimate the size of the sum and define a packed decimal work field that will contain it. The first field that will participate in the sum can be ZAPed into the work field. All other fields that contribute to the sum will be added using the AP instruction. The following example computes the sum of 3 packed decimal fields.

ZAP SUM,FIELD1

AP SUM,FIELD2

AP SUM,FIELD3

...

FIELD1 DS PL3



FIELD2 DS PL3

FIELD3 DS PL3

SUM DS PL7
Notice that SUM was uninitialized, but was zeroed out by the ZAP operation prior to the addition of FIELD1. The size of SUM is somewhat arbitrary and could vary based on our knowledge of the data. In the code above we have avoided the cardinal error of choosing a field that is too small to hold the result - a packed length 7 field will hold the sum of any 3 packed length 3 fields.

Subtracting Packed Decimal Fields

The comments above about adding packed fields also apply when subtracting them. Use the SP instruction to perform the subtraction. The main error to avoid is not providing a field large enough to hold the final result. The code below will compute the difference of FIELDA AND FIELDB.


ZAP DIFFER,FIELDA

SP DIFFER,FIELDB

...

FIELDA DS PL3



FIELDB DS PL3

DIFFER DS PL4


The SP instruction is useful for zeroing out a packed decimal field. Subtracting a field from itself will accomplish this result.
SP DIFFER,DIFFER DIFFER = 0
Comparing Packed Decimal Fields
It is often necessary to compare two packed decimal fields and branch based on how the two fields compare to each other. In assembly language, packed fields can be compared using the CP instruction. This has the effect of setting the condition code. Branch instructions are then coded in order to test the condition code and branch accordingly. Consider the following example which leaves the “larger” of two packed fields in a field called “BIGGER”. First FIELDA is copied to BIGGER, then the fields are compared. A branch instruction, BNH, tests the condition code and a branch occurs to the label “THERE” if the condition code is “Not High”. In other words, a branch occurs if FIELDB is equal or less than FIELDA. On the other hand, if FIELDB is larger, the branch is not taken and execution continues with the ZAP which copies FIELDB over the previous value in BIGGER.
ZAP BIGGER,FIELDA ASSUME FIELDA >= FIELDB

CP FIELDB,FIELDA FIELDB > FIELDA?

BNH THERE BRANCH IF EQUAL OR LOW

ZAP BIGGER,FIELDB CHANGE TO THE LARGER VALUE

THERE EQU *
It is a beginner’s mistake to compare packed fields with the CLC instruction. The compare logical character instruction was not designed to accommodate packed decimal data. The following code illustrates some of the problems that can occur.
CP AFIELD,BFIELD CONDITION CODE = EQUAL

CLC AFIELD,BFIELD CONDITION CODE = HIGH

CLC SHORTNO,LONGNO CONDITION CODE = HIGH

...


AFIELD DC X’12345C’ AFIELD = +12345

BFIELD DC X’12345A’ AFIELD = +12345

SHORTNO DC X’123C’

LONGNO DC X’0000123C’


Using CP in the first line, the fields are properly compared as equal packed decimal fields. ( Remember that C and A are valid plus signs for packed decimal data.) The first CLC instruction sets the condition code to “high” when it compares the third bytes of AFIELD and BFIELD. As character data, X’5C’ is higher than X’5A’. The second CLC illustrates another problem with using CLC. In this case, the condition code is set to high when comparing the first bytes as character data. In fact, the fields are equal when treated as packed decimal fields.
Multiplying Packed Decimal Fields
The MP mnemonic is used for multiplying packed decimal fields. This instruction contains two operands which are multiplied; the product is copied to the first operand, destroying the original contents. The following code illustrates how to multiply two fields together.
ZAP PRODUCT,FIELD1

MP PRODUCT,FIELD2

...

FIELD1 DS PL5



FIELD2 DS PL3

PRODUCT DS PL8


When planning to multiply two fields, in this case FIELD1 and FIELD2, you must plan for a field which is large enough to hold the product. The rule of thumb is that the length of the product field should be at least as large as the size of the multiplier length plus the multiplicand length. In the example above we compute the product length to be 5 + 3 = 8. The first step is to copy the multiplicand to the work field with the ZAP instruction. The operation is then completed by executing the MP instruction.
While Operand 1 ( containing the multiplicand ) can be as large as 16 bytes, Operand 2 (containing the multiplier ) is limited to a maximum of 8 bytes.
The MP instruction will cause your program to abend if there are not enough leading 0’s in the multiplicand prior to multiplication. The rule is that, prior to multiplying, there must be at least as many bytes of leading 0’s in the multiplicand as there are bytes in the multiplier. Consider the following example,
MP PRODUCT,FIELDB ABEND!

...


PRODUCT DC X’00001234567C’

FIELDB DC X’00887C’


The multiply instruction above causes an interrupt and the program abends because the multiplicand contains only 2 bytes of leading 0’s, while the multiplier is 3 bytes in length.
Dividing Packed Decimal Fields
Use the DP mnemonic for the division of packed decimal fields. Initially, Operand 1 is initialized with the dividend and the divisor occupies Operand 2. After the divide operation, Operand 1 contains the quotient, followed immediately by the remainder. Here is an example division which computes X / Y..
ZAP WORK,X INITIALIZE WITH THE DIVIDEND

DP WORK,Y Y IS THE DIVISOR

...

WORK DS 0CL8 GROUP FIELD



QUOT DS PL5 QUOTIENT OF X / Y

REM DS PL3 REMAINDER OF X / Y

X DS PL5

Y DS PL3
You must plan the size of each work area when dividing. In the example above, we are dividing a packed length 5 field by a packed length 3 field. The work area in which the division will occur must be large enough to contain a quotient and a remainder. How big could the quotient become? Since we are performing integer arithmetic, the quotient could be the same size as the dividend (consider division by 1). How big could the remainder become? The largest remainder is always one less than the divisor, but the field size of the remainder might be just as large as the divisor. Because of these considerations, the work area size should be at least as large as the size of the dividend plus the size of the divisor. In the code above, we made WORK eight bytes since the dividend was 5 bytes and the divisor was 3.


The first step was to ZAP the work area with the dividend, and then divide by Y. Suppose X initially contains X’000012356C’ and Y contains X’00100C’. After the division, WORK will contain X’000000123C00056C’. Notice that WORK is no longer packed, but contains two packed fields. It is a common error to reference the work area as a packed field after the division. This is a mistake which causes the program to abend.
Using the definition of WORK above, the following division would produce an error, eventually.
ZAP WORK,=P’123456’

DP WORK,=PL2’100’


The problem arises because the remainder’s size is completely determined by the divisor’s size. Since we divided by a 2 byte field, the remainder will occupy 2 bytes of WORK and the quotient fills the other 6 bytes. After the division, WORK contains X’00000001234C056C’, but the field definitions of QUOT AND REM do not match these results. A future reference to either of these fields as a packed decimal value will cause an abend.


Shifting Packed Decimal Fields
Since decimal points are not stored internally for packed decimal fields, and since packed decimal arithmetic is integer arithmetic, it is necessary for an assembler programmer to shift fields left and right in order to obtain the precision required for most calculations. This is accomplished with the shift and round pack instruction which has mnemonic SRP. ( Some shifts can be completed using the MVO instruction, but SRP is easier to use and offers more flexibility.) Using SRP, a packed decimal field can be shifted left or right while leaving the sign digit fixed. For right shifts, digits are lost on the right and 0’s fill in for digits which are shifted out on the left. For left shifts, leading 0’s are lost on the left and 0’s fill in for digits shifted out on the right.
The instruction has three operands: Operand 1 is the field that will be shifted, Operand 2 is the shift factor, and Operand 3 is a rounding factor for right shifts. The shift factor is a 6-bit 2’s complement integer that we will represent as a decimal integer between 1 and 31 for left shifts, and as 64 - n for right shifts of n digits. Operand 3, the rounding factor, is an integer from 0 to 9 that is added to the leftmost digit which is shifted out during a right shift. Any carry is propagated through the rest of Operand 1. Consider the following example.
SRP P,3,5 P = X’000123000C’

SRP Q,64-3,5 Q = X’0000010C’

...

P DC PL5’123’ P = X’000000123C’



Q DC PL4’9876 Q = X’0009876C’
In the first SRP, the shift factor of 3 indicates a left shift by 3 digits. Three digits are lost on the left and 3 zero digits are shifted in on the right. This shift is logically equivalent to multiplying by 1000. In the second SRP, the shift factor of 64 - 3 indicates a right shift by 3 digits. The 8, 7, and 6 are shifted off. Before shifting off the 8 which is the leftmost digit, the rounding factor of 5 is added to the contents of P. This addition causes a carry and creates the number 103 which is shifted, leaving a value of 10 in P.
Shifting is commonly used when working with integers that contain decimal points. Consider the problem of multiplying S and T, and leaving a product that contains 1 digit to the right of the decimal point. Remember that the machine does not store decimal points internally for packed decimal fields.
ZAP PRODUCT,S INITIALIZE THE MULTIPLICAND

MP PRODUCT,T ...2 DIGITS TO RIGHT OF DECIMAL PT

SRP PRODUCT,64-1,0 REMOVE ONE DIGIT WITHOUT ROUNDING

...


S DC PL3’1234.5’ S = X’12345C’ NO DECIMAL PT

T DC PL2’10.0’ T = X’100C’ NO DECIMAL PT

PRODUCT DS PL5
First the multiplicand is ZAPed into a 5 byte field called PRODUCT which is large enough to hold the product of S and T. The multiplication leaves PRODUCT with 2 digits to the right of the decimal point ( PRODUCT = X’001234500C’). The SRP shift out the rightmost digit leaving PRODUCT = X’000123450C’. This result could be edited using ED or EDMK and the decimal point could be inserted for printed output.

Arithmetic on packed decimal fields that “contain” decimal points requires some careful thought on the programmer’s part. Consider dividing 123.4 by 2.1 using integer arithmetic. Assume that after the division we would like the quotient to contain 1 decimal digit to the right of the decimal point. We illustrate two possible divisions below.


Keep in mind that decimal points are not stored internally. The first division illustrates dividing 2.1 into 123.4 . Since we are working with integers, this is equivalent to dividing 21 into 1234. The result is 58 and contains no decimal point. This will not give us the precision we demand in the quotient. In the second division, by shifting the dividend to the left by one digit (bringing in a 0 on the right), we are effectively dividing 21 into 12340, and producing a quotient of 587 which could be edited to 58.7 for printing.


The code below illustrates how the above division might appear in assembly language.
ZAP WORK,M M IS THE DIVIDEND

SRP WORK,1,0 M NEEDS MORE PRECISION

DP WORK,N N IS THE DIVISOR

MVC QUOTOUT,EDWD EDIT WORD GOES TO OUTPUT AREA

ED QUOTOUT,QUOT PREPARE QUOTIENT FOR PRINTING

...


M DC PL4’123.4’ M = X’0001234C’ (NO DECIMAL PT)

N DC PL2’2.1’ N = X’021C’ (NO DECIMAL PT)

WORK DS 0CL6 WORK FIELD FOR DIVISION

QUOT DS PL4 QUOTIENT

REM DS PL2 REMAINDER

QUOTOUT DS CL9

EDWD DC X’402020202021204B20’
The work area field was designed as 6 bytes since the dividend was 4 bytes and the divisor was 2 bytes. The dividend was moved to the work area and then shifted left for more precision. After the division, QUOT has the answer we would like to print. An edit word is created which matches the 4 byte packed field QUOT, and moved to an output area. QUOT is then edited into the output area. ( See ED for details on editing.)
Next we consider the problem of generating an answer that is rounded to a specified precision. Suppose we are going to divide 1234.56 by 2.1 and we would like to compute the quotient rounded to two decimal places to the right of the decimal point. If we simply divide, the quotient would have 1 digit to the right of the decimal point. In order to finish with a quotient that has two digits rounded, we must generate 3 digits to the right of the decimal point before we divide. We can achieve this precision by shifting to the left by 2 digits before dividing. The following code illustrates this idea.
ZAP WORK,X PREPARE THE DIVIDEND

SRP WORK,2,0 SHIFT IN 2 0’S ON THE RIGHT

DP WORK,Y Y IS THE DIVISOR

SRP QUOT,64-1,5 SHIFT RIGHT BY 1 AND ROUND

...

X DC PL4’1234.56’ X = X’0123456C’ (NO DECIMAL PT)



Y DC PL2’2.1’ Y = X’021C’ (NO DECIMAL PT)

WORK DS 0CL7 WORK AREA FOR DIVISION

QUOT DS PL5 QUOTIENT

REM DS PL2 REMAINDER


Why was WORK created as a 7 byte field when X contained 4 bytes and Y contained 2 bytes? The reason is that after moving X to WORK, we shifted it to the left by 2 digits, effectively making it a 5 byte field. Making WORK a 7 byte field insures we have enough room in the work area for the division. Since we divided by a 2 byte field, the remainder has 2 bytes and the rest of the work area is the quotient.
As a final example of handling decimal points, consider the problem of computing M times N and dividing the result by P. We would like the final answer to have 2 decimals to the right of the decimal point, rounded. The declarations of M, N, and P are listed below with the comments indicating the precision in each field.
M DS PL4 99999.99

N DS PL3 9999.9

P DS PL2 99.9
If we simply multiply M and N, the product will have 3 decimal places to the right of the decimal point. Dividing by P would reduce the number to 2. In order to finish with an answer that has 2 decimals rounded, we need 3 digits before shifting. That means the product must be shifted to the left by 1 digit before the division. The following code could be used.
ZAP WORK,M M IS THE MULTIPLICAND

MP WORK,N COMPUTE THE PRODUCT

SRP WORK,1,0 SHIFT IN A ZERO ON THE RIGHT

DP WORK,P QUOTIENT WILL HAVE 3 DECIMAL PLACES

SRP QUOT,64-1,5 ROUND QUOTIENT BACK TO 2 DIGITS

....


WORK DS 0CL10 WORK AREA

QUOT DS PL8 QUOTIENT



REM DS PL2 REMAINDER
Again, shifting the product left means the work area needs to be adjusted by one byte.
Yüklə 37,86 Kb.

Dostları ilə paylaş:




Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur ©genderi.org 2024
rəhbərliyinə müraciət

    Ana səhifə