[mpich-discuss] MPI_TYPE_MAX limit using MPT
Michael Raymond
mraymond at sgi.com
Fri Oct 15 06:50:26 CDT 2010
That's correct. In MPT every element of a complex type like a struct
or indexed type ends up taking up a type in the global table. There are
good speed reasons for this but the downside is that you sometimes need
to preallocate a ton of space for datatypes. This is remedied to some
extent in MPT 2.03. If you have any more questions about MPT you can
contact me directly.
Wei-keng Liao wrote:
> Hi, Max,
>
> I am cc this to mpich discuss group. Hope people there can provide more thorough
> answers.
>
> I have not run jobs on a SGI machine for a long time, but I remember I have to
> increase MPI_TYPE_MAX in order to run an MPI-IO job. In collective I/O, ROMIO
> uses MPI derived data type for exchanging I/O requests among processes. So,
> the more processes a job runs, the bigger number MPI_TYPE_MAX should be set.
>
> Wei-keng
>
> On Oct 6, 2010, at 11:29 AM, Maxwell Kelley wrote:
>
>> Hello,
>>
>> When running some I/O tests using MPT 1.25, I get the error
>>
>> MPI has run out of internal datatype entries.
>> Please set the environment variable MPI_TYPE_MAX for additional space.
>> The current value of MPI_TYPE_MAX is 8192
>>
>> Is this normal? Setting MPI_TYPE_MAX to 65536 simply allowed more I/O to be performed before the error appears. The limit is reached more quickly using more processors. Assuming that this is a case of types not being freed after use, should I just set this limit high enough that it will never be exceeded during a 12-hour batch job?
>>
>> -Max
>>
>
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
--
Michael A. Raymond
Message Passing Toolkit Team
Silicon Graphics Inc
(651) 683-3434
More information about the mpich-discuss
mailing list