I am trying to optimize a geometry at the CCSD(T) level with numerical<br>gradient using the MPPX parallel of MOLPRO 2006.1 (patch level 60). It<br>has 4 variables and needs 8 displacements. I used 8 processors on 4<br>dual-processor nodes and it produced 8 log files. After inspection of
<br>the log files, I found it didn't distribute the displacements in the<br>desired equal way. Instead they got 2, 1, 1, 1, 0, 1, 0, 2<br>displacements in the order of the processes respectively. Why two<br>processes got no work to do, while the other two got double amount of
<br>the work-load? Should I use more processors than the number of displacements? Thanks.<br><br>Best regards,<br clear="all"><br>-- <br>Shenggang Li