[mpich-discuss] The max amount of data each process can send in MPI_Alltoallv

[jt.meng]

Hi, Huiwei, 
       I think the best solution for your question is cuting the data buffers into smaller sizes.   However, This may be not easy. 
       According to my experience, the suggestions are 
           1.  Do not use  MPI_Alltoallv, and use nonblocking communication insted, such as MPI_send and MPI_irecv.  
           2.  In solution 1, each data buffer can not exceeds 2GB too,   If the data is too large,  I think we should develop a class encapsulated with the data buffer cuting method.
      
      However, using int as the parameters' type in MPI communication primitives is not reasonable.  As we know, in x86_64 machine,  sizeof(int)=4, however sizeof(long) and sizeof(long long int)=8, we should support the new parameter type of long and long long int.   




jt.meng

From: Huiwei Lu
Date: 2013-12-14 07:18
To: discuss
Subject: [mpich-discuss] The max amount of data each process can send in MPI_Alltoallv
Hi, folks,


What’s the max amount of data each process can send in MPI_Alltoallv? http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Alltoallv.html


As I understand, the input parameter of sendcnts is an array of int, so the amount of data each process can send is limited to MAX_INT (approximately 2G).


What if I want to use MPI_Alltoallv-like MPI routines to send a buffer of data that exceeds 2GB? Does it need to cut the data buffers to smaller sizes? Is there an easy way to do this? The reason I am asking is one of my distributed sort algorithm uses MPI_Alltoallv to distribute data after partitioning; it crashed when I tried to send in each process more than 2GB data.


Thanks.
--
Huiwei Lu
http://www.mcs.anl.gov/~huiweilu/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20131214/da04eb8f/attachment.html>