sists in exponent addition and fraction multiplica
tion. The operands are prenormalized, and the
less 64, is used as the characteristic of the intermedi
ate product.
The product of the fractions is developed such
that the result has the exact fraction product truncat
ed to the proper result-fraction length. When the
result is normalized without requiring any postnor
malization, the intermediate-product fraction is trun
cated to
intermediate-product characteristic becomes the
final product characteristic. When the intermediate
product
ed left one digit position, bringing the contents of
the guard-digit position into the low-order position
of the result fraction, and the intermediate-product
characteristic is reduced by one. The intermediate
product fraction is subsequently truncated to the
result-fraction length.
For MER and ME, the multiplier and multipli
cand have six-digit fractions, and the product frac
tion has the full 14 digits of the long format, with the
two low-order fraction digits always zero. For MDR
and MD, the multiplier and multiplicand fractions
have 14 digits, and the result product fraction is
truncated to 14 digits. For MXDR and MXD, the
multiplier
with the multiplicand occupying the high-order part
of the first
tains 28 digits and is an exact product of the operand
fractions. For MXR, the multiplier and multiplicand
fractions have 28 digits, and the result product frac
tion is truncated to 28 digits.
The sign of the product is determined by the rules
of algebra, unless all digits of the product fraction
are zero,
An exponent-overflow exception is recognized
when the characteristic of the normalized product
exceeds 127 and the fraction of the product is not
zero. The operation is completed by making the
characteristic 128 less than the correct value. If, for
extended results, the low-order characteristic also
exceeds 127, it, too, is decreased by 128. The result
is normalized, and the sign and fraction remain cor
rect. A program interruption for exponent overflow
then occurs.
Exponent overflow is not recognized if the
intermediate-product characteristic exceeds 127 but
is brought within range by normalization.
An exponent-underflow exception exists when the
characteristic of the normalized product is less than
zero and the fraction of the product is not zero. If
the exponent-underflow mask bit is one, the opera-
168 System/370 Principles of Operation
tion is completed by making the characteristic 128
larger than the correct value, and a program inter
ruption for exponent underflow occurs. The result is
normalized, and the sign and fraction remain correct.
If the exponent-underflow mask bit is zero, program
interruption does not take
tion is completed by making the product a true zero.
For extended results, exponent underflow is not
recognized when the low-order characteristic is less
than zero but the high-order characteristic is zero or
above.
Exponent underflow is not recognized when the
characteristic of an operand becomes less than zero
during pre normalization, but the characteristic of the
normalized product is within range.
When either or both operand fractions are zero,
the result is made a true zero, and no exceptions are
recognized.
The Rl field for MER, ME, MDR, and MD, and
the R2 field for MER, MDR, and MXDR must des
ignate register 0, 2, 4, or 6. The Rl field for MXDR,
MXD, and MXR, and the R2 field for MXR must
designate register
exception is recognized.
Com/ition Code:
The code remains unchanged.
Program Exceptions:
installed, or, for MXDR, MXD, and MXR, if
the extended-precision floating-point feature is
not installed)
Access (fetch, operand 2 of ME, MD, and MXD
only)
Specification
Exponent
multiplication does not affect the value of the prod
uct.
Store
[RX,