Hi

In the local cluster, the processors are shared by nodes. Each node should
present four processors, so I should run 4 processes by node. I ran the
transiesta calculation with 8 processors, then I used two nodes. The
calculation stops with an memory error. After, I run a calculation with
four processes and separate two on each node, the calculation runs normal.
Apparently, the first case uses more memory than the second.
This is my doubt, if the calculation is separated by node, the memory use
should be the same in these two cases?

Yoshi

Responder a