[mpich2-dev] [PATCH 1/1] Issue 4120:Read/write size error checking.

[James S Perrin]

Hi,

I think it was me who sent the original message "MPI I/O Reading 2gb or 
more of data" 27 Aug 2007. I assigned count_max = 2^31/size_of(type), 
any message larger than count_max is chunked.

Regards
James

Robert Latham wrote:
> On Thu, Mar 20, 2008 at 04:31:44PM -0500, Bob Cernohous wrote:
>> I'm not sure about this one.  I'm bothered when i see
>> count*datatype_size in the code and I imagine a >= 2G total
>> size that causes odd problems later.  So I added error
>> checking up front.  (This is different from writing/reading
>> data at >= 2G offsets in a file which we're trying to support.)
>>
>> For you review and approval.  I've kept the patch separate so
>> I can easily add or remove it.
> 
> I can't find the email, but I know we've had this come up before.  We
> say "well if you're working with such large amounts of data, you're
> doing pretty well".   Seems exactly right that we should error early
> on with an informative message.
> 
> You get a +1 from me.   
> 


-- 
------------------------------------------------------------------------
   James S. Perrin
   Visualization

   Research Computing Services
   The University of Manchester
   Kilburn Building, Oxford Road
   Manchester, M13 9PL

   t: +44 (0) 161 275 6945
   e: james.perrin at manchester.ac.uk
   w: www.manchester.ac.uk/researchcomputing
------------------------------------------------------------------------
  "The test of intellect is the refusal to belabour the obvious"
  - Alfred Bester
------------------------------------------------------------------------