Fix number of nodes estimation. (#5)

This is a temporary fix that let's correctly estimate the number
of nodes for the CSCS supercomputers. The issue here is different nodes
have more or less the same amount of RAM, so the number of CPUs
can't be used as a criterion to select the number of nodes.

Once the following issue is fixed: aiidateam/aiida-core#4297,
we can implement a better mechanism to auto-select the number of nodes.

Co-authored-by: Carlo Pignedoli <carlo.pignedoli@empa.ch>

aiida_nanotech_empa/workflows/nanoribbon.py

@@ -383,9 +383,10 @@ def _submit_pw_calc(self,
             1 + int(nkpoints * 2.4 /
                     builder.code.computer.get_default_mpiprocs_per_machine()),
             int(5))
-        nnodes = (1 + int(
-            natoms * 0.2 /
-            builder.code.computer.get_default_mpiprocs_per_machine())) * npools
+        #nnodes = (1 + int(
+        #    natoms * 0.2 /
+        #    builder.code.computer.get_default_mpiprocs_per_machine())) * npools
+        nnodes = (1 + natoms/60) * npools
 
         builder.metadata.label = label
         builder.metadata.options = {