In short-precision, the low-order halves of the Hoat­
ing-point registers are ignored and remain unchanged.
The SUBTRACT NORMALIZED is similar to ADD NORMAL­
IZED, except that the sign of the second operand is
inverted before addition.
The sign of the difference is derived by the rules of
algebra. The sign of a difference with zero result
fraction is always positive.
Resulting Condition Code:
o Result fraction is zero
1 Result is less than zero
2 Result is greater than zero
3 Result exponent overHows
Program Interruptions: Operation (if Hoating-point feature is not in-
stalled)
Addressing (SD and SE only)
Specification
Significance
Exponent overHow
Exponent underHow
Subtract Unnormalized SUR RR. (Short Operands)
3F
78 11 12 15 SU RX (Short Operands)
7F
7 8 11 12 1516 1920 SWR (Long Operands)
2F
7 8 11 12 15 SW RX (Long Operands)
6F
7 8 11 12 1516 1920 D2
31
31
The second operand is subtracted from the first op­
erand, and the unnormalized difference is placed in
the first operand location.
In short-precision, the low-order halves of the Hoat­
ing-point register are ignored and remain unchanged.
46
The SUBTRACT UNNORMALIZED is similar to ADD UN­ NORMALIZED, except for the inversion of the sign of the
second operand before addition.
The sign of the difference is derived by the rules of
algebra. The sign of a difference with zero result
fraction is always positive.
Resulting Condition Code:
o Result fraction is zero
1 Result is less than zero
2 Result is greater than zero
3 Result exponent overHows
Program Interruptions: Operation (if Hoating-point feature is not in-
stalled)
Addressing (sw and su only)
Specification
Significance
Exponent overHow Compare CER RR (Short Operands)
39
7 8 11 12 15
CE RX (Short Operands) I 79 Rl I X
2 I B2 0 7 8 11 12 1516 1920 CDR RR (Long Operands) I 29 Rl I R2 I 0 78 11 12 15
CD RX (Long Operands) I 69 Rl I X
2 I B2 0 7 8 11 12 1516 1920 31
31
The first operand is compared with the second op­
erand, and the condition code indicates the result.
In short-precision, the low-order halves of the Hoat­
ing-point registers are ignored.
Comparison is algebraic, taking into account the
sign, fraction, and exponent of each number. An expo­
nent inequality is not decisive for magnitude determi­
nation since the fractions may have different numbers
of leading zeros. An equality is established by follow­
ing the rules for normalized Hoating-point subtraction.
When the intermediate sum, including a possible
guard digit, is zero, the operands are equal. Neither
operand is changed as a result of the operation.
Exponent overflow, exponent underflow, or lost sig­
nificance cannot occur.
Resulting Condition Code:
o Operands are equal
1 First operand is low
2 First operand is high
3
Program Interruptions:
Operation (if floating-point feature is not in­
stalled)
Addressing (CD and CE only)
Significance
Programming Note
Numbers with zero fraction compare equal even when
they differ in sign or characteristic. Halve HER RR (Short Operands)
34
7 8 11 12 15
HDR RR (Long Operands)
24
7 8 11 12 15
The second operand is divided by two, and the
quotient is placed in the first operand location.
In short-precision, the low-order half of the result
register remains unchanged.
The operation shifts the fraction right one bit; the
sign and characteristic are not changed. No normaliza­
tion or test for zero fraction takes place.
Condition Code: The code remains unchanged.
Program Interruptions: Operation (if floating-point feature is not in­
stalled)
Specification
Programming Note
The halve operation differs from a divide operation
with the number two as divisor in the absence of pre­
normalization and postnormalization and in the ab­
sence of a zero-fraction test. Multiply MER RR (Short Operands) I 3C R] I R2 0 7 8 11 12 15
ME RX (Short Operands)
7C R] I X
2 I B2
7 8 11 12 1516 1920 31
MDR RR (Long Operands) I 2C R] I R2 I 0 7 8 11 12 15
MD RX (Long Operands) I 6C R] I X
2 I B2 0 7 8 11 12 1516 1920 31
The normalized product of multiplier (the second op­
erand) and multiplicand (the first operand) replaces
the multiplicand.
The multiplication of two floating-point numbers
consists of a characteristic addition and a fraction
multiplication. The sum of the characteristics less 64 is
used as the characteristic of an intermediate product.
The sign of the product is determined by the rules of
algebra.
The product fraction is normalized by prenormaliz­
ing the operands and postnormalizing the intermediate
product, if necessary. The intermediate product char­
acteristic is reduced by the number of left-shifts. For
long operands, the intermediate product fraction is
truncated before the left-shifting, if any. For short
operands (six-digit fractions), the product fraction has
the full 14 digits of the long format, and the two low­
order fraction digits are accordingly always zero.
Exponent overflow occurs if the final product char­
acteristic exceeds 127. The operation is terminated,
and a program interruption occurs. The overflow ex­
ception does not occur for an intermediate product
characteristic exceeding 127 when the final character­
istic is brought within range because of normalization.
Exponent underflow occurs if the final product char-
Floating-Point Arithmetic 47
Previous Page Next Page