Texas Instruments TMS320C3x - Manual (Page 144)

Rounding (RND Instruction)
5-39 Data Formats and Floating-Point Operation
5.8 Rounding (RND Instruction)
The RND instruction rounds a number from the extended-precision floating-
point format to the single-precision floating-point format. Rounding is similar to
floating-point addition. Given the number
a to be rounded, the following opera-
tion is performed first.
c =
α
( man)
×
2
α
( exp) + (1
×
2
α
( exp) –24 )
Next, a conversion from extended-precision floating-point to single-precision
floating-point format is performed. Given the extended-precision floating-point
value, the rounding, rnd( ), is performed as shown in Figure 5–19.
Note:
RND, src, dst—where (src) = 0—does not set the 0 conditions flag (bit 2 in
the status register). Instead, it sets the underflow condition flag (bit 4 in the
status register). When required, check for the underflow condition instead of
the 0 condition.