On Monday, Nov 10, 2003, at 11:03 US/Pacific, piclist@XARGS.COM wrote:
> I assume you meant upper 8 bits?
>
> unsigned char c;
> int i;
>
> c = i >> 8;
>
> generates optimal code.  So will a union, but the above is more
> portable.
>
>> right now I'm doing it like this:
>> (char)(number/256)
>
> If "number" is an unsigned int, it will also generate optimal code;
> i.e., it doesn't actually do a division.
>

Whether either of these statements will generate 'optimal' code is
rather compiler dependent.  Sure, it'd be nice if most c compilers
optimized this, but I'm not sure that it's true.  I was looking at the
jal source code recently, and was rather shocked at its "lazy"
implementation of multiply and divide, for instance.  Look at the asm
file generated, and
see what it actually does...

BillW

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.