actually from within HPC Cluster manager, you still need to start
mpiexec with proper parameters, rather than Firefly itself.
You need to pass the file with the list of allocated nodes
(or just take them from the corresponding environment variable)
to the mpiexec and direct it to run Firefly on these nodes.
Hope this helps,
On Fri Nov 20 '09 3:52pm, Thomas wrote
>We are tying to start PCGAMESS on a Windows HPC Cluster 2008. It consists of 3 nodes with a total of 10 cores (Intel Core 2). PCGAMESS works well at each node as well at the master node, if we start it in the cmd command prompt (e.g. 'pcgamess ľo testout.out').
>It does not work, if we try to start the program in the HPC Cluster Manager (also command line e.g. 'pcgamess ľo testout.out'). We use as descript the MPIBIND.MSMPI.dll as mpibind.dll. We always get a failled process with an error code 0 in the HPC Manager Software.
>Does anyone have experience in using PCGAMMESS in Windows HPC Clusters or any helpful remarks?
>Thanks and regards,
[ This message was edited on Sat Nov 21 '09 at 12:26pm by the author ]