Re: [Help-glpk] bigmul in glblib05.c

[Michael Hennebry]

On Wed, 6 Jan 2010, Andrew Makhorin wrote:

> > These codes are replaced with
> >
> > unsigned short c = (unsigned short)( i & USHRT_MAX);
> >
> > The modified codes can produce 13.75 for vpm2.mps of MIPLIB 2003.
>
> I am not sure why you are changing the code. Doesn't the original
> version work for you? The glpk code is highly portable, and there is
> no need to "improve" its portability in such way.
>
> The statement:
>
>   unsigned short c = (unsigned short)( i & USHRT_MAX);
>
> is absolutely equivalent to:
>
>   unsigned short c = (unsigned short)i;

Actually it's not.
The result of the latter statement can be determined
from the value of i and the value of USHRT_MAX.
If i is negative, the result of the former also
depends on the representation of negative numbers.
I'd go with the latter.
Failing that, I'd put in a test for twos complement:
unsigned short c = (unsigned short)( (i & USHRT_MAX) + (1-1/(-1==1+~1)));

--
Michael   address@hidden
"Pessimist: The glass is half empty.
Optimist:   The glass is half full.
Engineer:   The glass is twice as big as it needs to be."