🚀 go-pugleaf

In article <71790001@hplvec.LVLD.HP.COM> boyne@hplvec.LVLD.HP.COM (Art Boyne) writes:
>In comp.std.c, dik@cwi.nl (Dik T. Winter) writes:
>
>>   What it comes down to is that << is not a mathematical operation, there
>>   is nothing in mathematics that tells us that 16384 << 1 equals 32768.
>>   So what is the meaning of:
>
>I'm not so sure about that.  Conceptually, the common-usage interpretation
>of the operation a << b can be represented mathematically as a*(2^b).

Umm, not necessarily on a machine using sign-magnitude representation.
Not on a twos-complement machine if the shift ends up changing the
value of the leftmost bit, even if the register does not overflow.  Not
on a ones-complement machine (are there still any of those?) if a is
negative, when the effective operation becomes a*(2^b)-1.

>Given that potential implementation, it is easy to imagine that INT_MAX << 1
>would generate overflow on such a machine.  Now, considering that the ANSI
>standard in general places as few restrictions as possible on implementations
>so as to maximize performance, I'm not sure that there is anything in the
>standard that would judge this implementation as non-conforming.

Well, OK, sod the standard for a lark, then. :-)  Let's look at K&R 1:
'The value of E1<<E2 is E1 (interpreted as a bit pattern) left-shifted
E2 bits.'  I would claim that any 'old' C programmer will expect that
behaviour.  I (and I suspect most 'oldies') would neither buy nor
recommend any C compiler which generated an exception condition for any
'valid' shift  (i.e. E2 positive, and less than the length of E1).  If
I want multiplication by a power of two with an exception on overflow,
I will WRITE a*(2^b).  Maybe market forces will win.

-- 
Paul Smee, Computing Service, University of Bristol, Bristol BS8 1UD, UK
 P.Smee@bristol.ac.uk - ..!uunet!uknet!bsmail!p.smee - Tel +44 272 303132