Re: alignment boundary

[Bob Friesenhahn]

On Sat, 12 May 2007, Ralf Wildenhues wrote:
> > One approach is to use malloc() to allocate some memory and see what
> the alignment of the returned address is:
>
> Its alignment could be much larger than needed, no?

Yes, possibly.  But no less.

> If you have only a finite set of data types to check, then you can use
> AC_CHECK_ALIGNOF on (a union of) the types.

In that case be sure to test for the 128 bit integer and double types 
supported by some common CPUs. Given the lack of standardization for 
how to specify these extended types ('long long int'? 'long double'?) 
developing a configure test could be interesting.

Bob
======================================
Bob Friesenhahn
address@hidden, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/