[<<][staapl][>>][..]

signed/unsigned comparisons

two issues. are they the same or not, and what should the default be?

they are not the same:

     pos neg >
       * always true in signed
       * always false in unsigned

unsigned: carry
signed: sign of result (might overflow)

it's a bit silly, but i think it's time i admit i don't fully
understand it.. carry in addition is simple. carry in subtraction is
also not so difficult, since subtraction is addition with negative.

a carry on addition means overflow: the word's not big enough. simple.

but what is a carry on subtraction? let's isolate some cases.

            result carry sign overflow
10 3 -        7      1     0    0
3 10 -       -7      0     1    0

100 -100 -  -53      0     1    1
-100 100 -   53      1     0    1


http://en.wikipedia.org/wiki/Overflow_flag

  The overflow flag is usually computed as the xor of the carry into
  the sign bit and the carry out of the sign bit.

In other words: addition adds one extra bit to the representation. In
order to not have overflow, for unsigned addition/subtraction this bit
needs to be 0, and for signed addition/subtraction this needs to be
the same as the sign bit.

So, for a signed comparison, take the sign bit of the result, and
assume there is no overflow. For unsigned take the carry bit.




[Reply][About]
[<<][staapl][>>][..]