[mpich-discuss] MPI IO performance

[burlen]

I have made use of MPI-IO in my application. I have arrays stored one 
per file in fortran order and I have set the file view to a subarray and 
used  MPI_File_read_all to read that subarray using native mode. I have 
made some comparison against std::ifstream to read the same array with 
the same decomposition I am finding std::ifstream is a bit faster, some 
cases more than 40% faster. However, I had expected to see at worst 
equivalent performance.

Are there any issues I have to watch out for using MPI-IO? What could 
explain the results? What I can do to improve the situation?

My test array is 513Mb, I have tested on 2,4, and 8 nodes, with a single 
process per node. File system is nfs. For my tests I am using 
mpich2-1.0.8p1 and ch3:ssm.

Burlen