Analog Devices ADSP-BF53x Blackfin Reference page 81

UNROUNDED VALUE:
X X X X X X X X
ADD 1 AND CARRY:
. . . . . . .
ROUNDED VALUE:
X X X X X X X X
A0.X
Figure 2-7. Typical Unbiased Multiplier Rounding
The compensation to avoid net bias becomes visible when all lower 15 bits
are 0 and bit 15 is 1 (the midpoint value) as shown in
In Figure 2-8,
every rounding operation, but is evident only when the bit patterns shown
in the lower 16 bits of the next example are present.
ADSP-BF53x/BF56x Blackfin Processor Programming Reference
X X X X X X X X 0 0 1 0 0 1 0 1
. . . . . . . . . .
X X X X X X X X 0 0 1 0 0 1 1 0
bit 16 is forced to 0. This algorithm is employed on
A0
Computational Units
1 X X X X X X X X X X X X X X X
. . . . . . . 1 . .
0 X X X X X X X X X X X X X X X
A0.W
. . . . . . . . . .
Figure 2-7.
2-21
. . .