[mpich2-dev] Fw: 32-bit ROMIO: What is the best solution?

[Brian Smith]

I was probably over zealous in my statement. I know there are a lot of 
MPI_Aint == void * assumptions and some MPI_Aint = int * assumptions. 
There are probably no int = int * assumptions since, yes, a 64-bit 
compiler should find those. I'll let you know if we find any though. :)



Brian Smith
BlueGene MPI Development
IBM Rochester
Phone: 507 253 4717
smithbr at us.ibm.com




"Rajeev Thakur" <thakur at mcs.anl.gov> 
02/29/2008 05:42 PM

To
"'William Gropp'" <gropp at mcs.anl.gov>, Brian Smith/Rochester/IBM at IBMUS
cc
<mpich2-dev at mcs.anl.gov>, Bob Cernohous/Rochester/IBM at IBMUS
Subject
RE: [mpich2-dev] Fw: 32-bit ROMIO: What is the best solution?






I'd be surprised if there are any sizeof(int*) == sizeof(int) assumptions, 
because compilers on 64-bit systems would have warned about them. But 
there could be.
 
Rajeev

From: William Gropp [mailto:gropp at mcs.anl.gov] 
Sent: Friday, February 29, 2008 11:30 AM
To: Brian Smith
Cc: mpich2-dev at mcs.anl.gov; Bob Cernohous; owner-mpich2-dev at mcs.anl.gov
Subject: Re: [mpich2-dev] Fw: 32-bit ROMIO: What is the best solution?

Any place where sizeof(int*) == sizeof(int) is assumed is just broken - if 
its easy to separate those out, we'll just apply a fix for those, since 
that's needed for any platform and under all conditions. The 
sizeof(MPI_Aint) == sizeof(void*) assumption will require more care. 

Bill

On Feb 29, 2008, at 11:08 AM, Brian Smith wrote:


That's basically what we are doing. The problem is, there are a large 
number of places throughout the code (not just in ROMIO) where there are 
casts that assume MPI_Aint == sizeof(void*)==sizeof(int*)==sizeof(int), so 
we are walking through all that code and seeing what the casts are doing 
and fixing them as appropriate. 

William Gropp
Paul and Cynthia Saylor Professor of Computer Science
University of Illinois Urbana-Champaign


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mcs.anl.gov/mailman/private/mpich2-dev/attachments/20080229/2dfd2976/attachment.htm>