[mvapich-discuss] How to display all MPI traffic for a task?

[Thiago Quirino - NOAA Federal]

Thank you so much, Hari. I apologize for what seems to be now additional
work. I think I will manually try to print the message tag values right
before the MPI_*Send calls.

Again, thank you!
Thiago.

On Mon, Sep 29, 2014 at 12:52 PM, Hari Subramoni <subramoni.1 at osu.edu>
wrote:

> Dear Thiago,
>
> Currently MVAPICH2 does not have such a feature. We will explore some
> options to see if we can have a patch for this and get back to you.
>
> Regards,
> Hari.
>
> On Fri, Sep 26, 2014 at 10:17 AM, Thiago Quirino - NOAA Federal <
> thiago.quirino at noaa.gov> wrote:
>
>> Hi, folks.
>>
>> I am working on a numerical weather prediction model that is built on
>> MPI. There is a call to MPI_Iprobe in a subroutine that is looking for a
>> message with a specific tag from a specific host (i.e. neither of "
>> MPI_ANY_SOURCE" or "MPI_ANY_TAG" were used). MPI_Iprobe is returning
>> "true" when in fact the querying host should not have received a message.
>> Actually, the subsequent call to MPI_Recv fails because it expects a
>> payload of 12 bytes (3 integers), but 20 bytes were received (i.e. "Message
>> truncated" error).
>>
>> So I am trying to figure out where within the vast and complex code a
>> potential message with the specific tag and from a specific host is being
>> sent out, that is, a potential tag collision. I am wondering if there is a
>> debug flag or runtime variable that can be set for MVAPICH2 that will show
>> all MPI traffic for a task, for example, "Host 0 MPI_Send to Host 10 with
>> tag 20673", or something like that so that I can locate the bug.
>>
>> Thank you abundantly for any tips that you may have.
>>
>> All the best,
>> Thiago.
>>
>>
>> ---------------------------------------------------
>> Thiago Quirino, Ph.D.
>> NOAA Hurricane Research Division
>> Numerical Modeling Group
>> 4301 Rickenbacker Cswy.
>> Miami, FL 33149
>> P: 305-361-4337
>> C: 305-409-9587
>> E: Thiago.Quirino at noaa.gov
>> http://hwrf.aoml.noaa.gov
>> "The contents of this message are mine personally and do not necessarily
>> reflect any position of NOAA, its affiliates, or any branch of the federal
>> and local government."
>>
>> _______________________________________________
>> mvapich-discuss mailing list
>> mvapich-discuss at cse.ohio-state.edu
>> http://mailman.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20140929/9d2df51e/attachment.html>