pnetcdf I/O inefficiency problem

[萧 澍]

Hi,
This is Darin Shaw, a ordinary user of pnetcdf. Recently, I tackle a problem about the I/O, when I try to read variables from nc file with over 2k processors (every processor read a part of data), nc file reading efficiency is extremely slow, while reading efficiency is normal with less than 2k processors. I find that program halt in function ‘nfmpi_iget_vara_double’.
Do you meet the same kind of problem before or have any advice on how to solve it?
Thanks,
Darin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/parallel-netcdf/attachments/20190321/252dfb60/attachment.html>