d3d/archive/D3D11_3_FunctionalSpec.htm [24268:24413]: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Description: Component-wise divide. Operation: dest = src0 / src1

The following table shows the results obtained when executing the instruction with various classes of numbers, assuming that neither overflow or underflow occurs.

Beware of the two allowed implementations of divide: a/b and a*(1/b).

One outcome of this is there are exceptions to the table below for large denominator values (greater than 8.5070592e+37), where 1/denominator is a denorm. Since implementations may perform divide as a*(1/b), instead of a/b directly, and 1/[large value] is a denorm that could get flushed, some cases in the table would produce different results. For example (+/-)INF / (+/-)[value > 8.5070592e+37] may produce NaN on some implementations, but (+/-)INF on other implementations.

src0    src1-> -inf -F -1.0 -denorm -0 +0 +denorm +1.0 +F +inf NaN
-inf NaN +inf +inf +inf +inf -inf -inf -inf -inf NaN NaN
-F +0 +F -src0 +inf +inf -inf -inf src0 -F -0 NaN
-denorm +0 +0 +0 NaN NaN NaN NaN -0 -0 -0 NaN
-0 +0 +0 +0 NaN NaN NaN NaN -0 -0 -0 NaN
+0 -0 -0 -0 NaN NaN NaN NaN +0 +0 +0 NaN
+denorm -0 -0 -0 NaN NaN NaN NaN +0 +0 +0 NaN
+F -0 -F -src0 -inf -inf +inf +inf src0 +F +0 NaN
+inf NaN -inf -inf -inf -inf +inf +inf +inf +inf NaN NaN
NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN

F means finite-real number.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - d3d/archive/images/d3d11/D3D11_3_FunctionalSpec.htm [22069:22214]: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Description: Component-wise divide. Operation: dest = src0 / src1

The following table shows the results obtained when executing the instruction with various classes of numbers, assuming that neither overflow or underflow occurs.

Beware of the two allowed implementations of divide: a/b and a*(1/b).

One outcome of this is there are exceptions to the table below for large denominator values (greater than 8.5070592e+37), where 1/denominator is a denorm. Since implementations may perform divide as a*(1/b), instead of a/b directly, and 1/[large value] is a denorm that could get flushed, some cases in the table would produce different results. For example (+/-)INF / (+/-)[value > 8.5070592e+37] may produce NaN on some implementations, but (+/-)INF on other implementations.

src0    src1-> -inf -F -1.0 -denorm -0 +0 +denorm +1.0 +F +inf NaN
-inf NaN +inf +inf +inf +inf -inf -inf -inf -inf NaN NaN
-F +0 +F -src0 +inf +inf -inf -inf src0 -F -0 NaN
-denorm +0 +0 +0 NaN NaN NaN NaN -0 -0 -0 NaN
-0 +0 +0 +0 NaN NaN NaN NaN -0 -0 -0 NaN
+0 -0 -0 -0 NaN NaN NaN NaN +0 +0 +0 NaN
+denorm -0 -0 -0 NaN NaN NaN NaN +0 +0 +0 NaN
+F -0 -F -src0 -inf -inf +inf +inf src0 +F +0 NaN
+inf NaN -inf -inf -inf -inf +inf +inf +inf +inf NaN NaN
NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN

F means finite-real number.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -