X86 OF and CF calculaton

For sub and add instructions CF can be calculated using comparing operator for unsigned values:

sub_cf = (src1 <src2)

add_cf = (res < src1)

Note, that for add res < src1 if and only if res < src2 so it doesn't matter which argument to compare against the result.

Both CF and OF flags can be calculated knowing sign bits of the result and arguments:

sub_cf = (src1&src2&res) | ((~src1)&(src2|res))

sub_of = (src1^src2)&(src1^res)

add_cf = ((src1^src2)&(~res)) | (src1&src2)

add_of = (~(src1^src2))&(src1^res)

To assert first formula see that we can carry if the minuend is large, the subtrahend is large and the result is large. If the subtrahend or the result are small then carry is unnecessary. If the minuend is small then we should carry if either the subtrahend or the result is large.
The second formula says that signed subtraction overflows only if the minuend and the subtrahend have different sign and furthermore, the result has the sign different from the minuend.
Third formula uses the fact that unsigned addition generates carry only if both summand are large or if one of the summands is large and the result is not.
The last formula verifies that signed add overflows only if both summands have the same sign different from result's one.

PS as for the note, on ARM sub_cf is the same as on x86, but add_cf is inverted