Single-Precision Floating-Point Subtract
SUBSP
4-81
TMS320C67x Floating-Point Instruction Set
Notes:
1) If rounding is performed, the INEX bit is set.
2) If one source is SNaN or QNaN, the result is NaN_out. If either source
is SNaN, the INVAL bit is set also.
3) If both sources are +infinity or –infinity, the result is NaN_out and the
INVAL bit is set.
4) If one source is signed infinity and the other source is anything except
NaN or signed infinity of the same sign, the result is signed infinity and
the INFO bit is set.
5) If overflow occurs, the INEX and OVER bits are set and the results are
set as follows (LFPN is the largest floating-point number):
Overflow Output Rounding Mode
Result Sign
Nearest Even
Zero
+Infinity
–Infinity
+
+infinity
+LFPN
+infinity
+LFPN
–
–infinity
–LFPN
–LFPN
–infinity
6) If underflow occurs, the INEX and UNDER bits are set and the results
are set as follows (SPFN is the smallest floating-point number):
Underflow Output Rounding Mode
Result Sign
Nearest Even
Zero
+Infinity
–Infinity
+
+0
+0
+SFPN
+0
–
–0
–0
–0
–SFPN
7) If the sources are equal numbers of the same sign, the result is +0 unless
the rounding mode is –infinity, in which case the result is –0.
8) If the sources are both 0 with opposite signs or both denormalized with
opposite signs, the sign of the result is the same as the sign of
src1.
9) A signed denormalized source is treated as a signed 0 and the DENn bit
is set. If the other source is not NaN or signed infinity, the INEX bit is also
set.
