[mpich-devel] [mpich-ibm] Regression tests on mpich-3.1.2

Rob Latham robl at mcs.anl.gov
Fri Aug 8 14:45:29 CDT 2014



On 08/04/2014 09:12 AM, Rob Latham wrote:
>
>
> On 08/04/2014 08:43 AM, Junchao Zhang wrote:
>> Did you mean mpich-3.1.2,  instead of 1.3.2?  Also, if it is syntax
>> error, could you post the error message?
>
> Yeah, Will meant 3.1.2 (to confirm 'bigio' was first in the 3.1.2 release).
>
> the 'bigio' failure is when an i/o request overflows an integer type.
> Going to be a bit of work to fix that one because the integer type array
> is passed to  MPI_TYPE_CREATE_HINDEXED and a few other function
> expecting "array of pointer to ints".

OK, bigtype is fixed now and is under code review.

Jeff, I cribbed mightily from your BigMPI code, so thanks for putting it 
on github.

==rob


-- 
Rob Latham
Mathematics and Computer Science Division
Argonne National Lab, IL USA


More information about the devel mailing list