I have an assignment to do a logical shift n spots to the right. On my WinXP machine using Dev-C++ >> is a logical shift. On my school's Linux machine, or maybe there is a flag in the compiler?, it does an arithmetic shift. C is not taught in this class, just general programming. I have search dozens of C websites and they all just tell me that >> does a right shift. How do I tell C I want a logical shift or an arithmetic shift?
Thanks.
C - logical vs arithmetic right shift
Started by StubornAH, Sep 27 2005 06:24 AM
7 replies to this topic
Sponsor:
#3 Members - Reputation: 850
Posted 27 September 2005 - 06:27 AM
In general C compilers perform logical right shifts on unsigned data types and arithmetic right shifts (those that maintain the sign) on signed data types.
However it's something of a trick question since the compiler isn't required to support arithmetic ones at all and may choose to use logical ones exclusively.
It's also interesting to know that a division by two is not necessarily (and almost always in practice) equivalent to a two's complement arithmetic right shift. The division rounds towards zero (although it doesn't have to) while the shift does not.
That's actually a Java operator.
BTW does anyone know what the standard says about one's completement machines and arithmetic right shifts? You'd think that the natural implementation, unlike the two's complement version, would round towards zero instead.
However it's something of a trick question since the compiler isn't required to support arithmetic ones at all and may choose to use logical ones exclusively.
It's also interesting to know that a division by two is not necessarily (and almost always in practice) equivalent to a two's complement arithmetic right shift. The division rounds towards zero (although it doesn't have to) while the shift does not.
signed a = -10;
unsigned b = a;
a >>= 1; // aritmetic (probably)
b >>= 1; // logical
Quote:
Original post by Tinyn
>>>
That's actually a Java operator.
BTW does anyone know what the standard says about one's completement machines and arithmetic right shifts? You'd think that the natural implementation, unlike the two's complement version, would round towards zero instead.
#8 Members - Reputation: 850
Posted 27 September 2005 - 07:39 AM
Quote:None that I know of (aside IEEE floats which kind of uses it internally). They do introduce some interesting exceptions in the language however. So they're great when you want to say 'well.. that's technically incorrect' =)
Original post by ZQJ
Do any modern CPUs actually use one's complement?
I wouldn't be surprised if a few old beasts were still up and running (even actively developed for) though..






