Bug in intel instruction set ?: NEG gives wrong...

Here's some food for thought on this subject:

I'll use signed 8-bit, though this extends to any width (16, 32, 64, etc)
signed integers

NEG of 00000000 = 00000000
NEG of 11111111 = 00000001
NEG of 11111110 = 00000010

NEG of 10000010 = 01111110
NEG of 10000001 = 01111111
NEG of 10000000 = 10000000

Two values here have a NEG that is equal to itself.

For integer math, we know NEG of zero is always zero: -1 * 0 = 0

The other instance occurs when you're at the extreme negative value for the
width you are using. You can think of the computation in a few ways.
10000000 = -128 decimal. NEG of -128 dec = +128 dec, which would be
represented as 10000000 in binary. (The only reason we do not have a +128 in
signed 8-bit is because we have that defined as -128 already.)

Another way to look at it is this:

you have the following value ranges for signed 8-bit:

-127 decimal to -1 decimal
0
+1 decimal to +127 decimal
and a special overflow case: +/- 128

Never let your values reach this overflow case, and you'll be fine.
--
Bx.C / x87asm

...