F-12
GUIDELINES FOR WRITING SIMD FLOATING-POINT EXCEPTION
Note: For denormal encodings, see Table 9-2, Chapter 9, Programming with the Streaming SIMD Exten-
sions.
Table F-13. #D - Denormal Operand
Instruction
Condition
Masked ResponseUnmasked Response and
Exception Code
ADDPS
SUBPS
MULPS
DIVPS
SQRTPS
MAXPS
MINPS
CMPPS
ADDSS
SUBSS
MULSS
DIVSS
SQRTSS
MAXSS
MINSS
CMPSS
COMISS
UCOMISS
src1 = denormal or
src2 = denormal
#DE=1
res = result rounded to the
destination precision and
using the bounded
exponent, but only if no
unmasked post-
computation exception
occurs
src1, src2 unchanged,
#DE=1
(SQRT only has 1 src)
Table F-14. #0 - Numeric Overflow
InstructionCondition
Masked Response
Unmasked Response and
Exception Code
ADDPS
SUBPS
MULPS
DIVPS
ADDSS
SUBSS
MULSS
DIVSS
rounded result
> largest single-
precision finite
normal value
Rounding SignResult & Status Flags
res = (result calculated with
unbounded exponent and
rounded to the destination
precision) / 2
192
#OE=1
#PE=1 if the result is
inexact
To
nearest
+
-
#OE=1, #PE=1
res =
res =
Toward +
-
#OE=1, #PE=1
res = 1.11
1 * 2
127
res =
Toward +
-
#OE=1, #PE=1
res =
res = -1.11
1 * 2
127
Toward
0
+
-
#OE=1, #PE=1
res = 1.11
1 * 2
127
res = -1.11
1 * 2
127
?
+
?
?
?
?
+
?
+