Peter wrote:
> gcc does indeed do sign extended right shift of negative integers, with
> horrible results

Actually, doing a arithmetic shift on signed numbers and a logical shift on
unsigned sounds like the single most reasonable way to implement the C >>
operator on machines where arithmetic and logical shifts have the same cost.
This is also totally consistant with the C standard, so the results should
be neither horrible nor unexpected.

> According to gcc -1 >> 1 = -1 which is a little bit sick.

No, it is the most reasonable outcome and totally within the standard.  If
there is anything stupid going on here, it's someone writing C code not
expecting this as a possibility.

Let's say we're on a machine where these numbers are 16 bit integer for
simplicity.  Would you rather -1 >> 1 be 32767?  I think -1 makes more
sense, although both would actually be legal C.

> Also -5 >> 1 = -3 (should be -2).

You need to go look up bit shifts and the C standard.  -5 in binary is
..11011.  Shifting it right one bit yields ...11101, which is -3 if a
arithmetic right shift was used and 32765 if logical shift was used (again
using 16 bit words for simplicity).  I'd say -3 is rather more intuitive,
but no way should the answer be -2, which is a bit pattern of ...11110 and
would also violate the C standard (for good reason).

I don't like the way the C >> operator is defined, but in this case (-5 >>
1) being -3 is both correct and the most logical choice.

> One would expect the >> to shift
> bits, without sign extension,

Not if one had bothered to read the description of ">>".

> I would expect >> to *always* take the lhs argument as an unsigned
> integer and shift it right, and *not* sign extend it.

Wrong.  RTFM (where M is the C standard).


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist