[mpich2-dev] [PATCH 1/1] Issue 4120: Fix attr alignment check from MPI_Aint to long alignment

[Dave Goodell]

On Mar 24, 2008, at 4:12 PM, Jeff Parker wrote:

>
> On 03/23/2008 04:26 PM, William Gropp wrote:
> "This one is system-specific - some systems will require the  
> MPI_Aint alignment.   This should probably be abstracted out and  
> set based on the requirements of the system."
>
> Are you saying that on some systems, "long" is not the size of a  
> pointer, whereas "MPI_Aint" is?  Are there really such systems that  
> we need to be concerned about?  Standard C specifies that long and  
> pointers are the same size.

Do you have a C standard spec section to which you can point me that  
guarantees this?  I can't seem to find one.  It happens to be true on  
any platform where I've actually paid attention to the type widths,  
but I'm pretty sure it's not mandated by the standard.  I think all  
that is required is an ordering between the type sizes (sizeof(int) 
<=sizeof(long)) and that (sizeof(char)==sizeof(unsigned char)==1) and  
that certain types have certain minimum limits which may be greater  
on a given platform.

-Dave