QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: John on July 8, 2010, 03:56
-
Dear everyone,
1)In the new version, it seems that the electrodes formations are always changed with the changes of electrode copy length, and the left and right periodic lead parts are not same like 2008.10, how do understand it.
2) If I want to run the job in parallel on Windows XP(one machine with 4 CPUs), whether I need to install the mpich2, and how to assgin the CPUs(just now I use the trial license, does it support the parallel)
Thank you!
-
1) I strongly recommend reading the upgrade guide (http://quantumwise.com/documents/tutorials/UpgradeGuide_10.8/XHTML/), it describes this in detail.
2) Yes, you need to install MPICH2 and use "mpiexec" which it provides. Are these separate computers, or one machine with 4 CPUs? You can run in parallel with the demo license (provided it contains the feature "atkdftslave").
Support for parallel calculations on Windows is still a bit limited, there are some potential issues with conflicts with the license system. Let us know how it works out, and we'll try to assist.
-
Ok, thank you! My PC is 4 cores Windows XP operating system, now I just want to run the job in parallel, but it seems that there are some issues, I can not realize it. whether are there any introductions about the parallel calculation on windows XP? thanks!
-
For a Windows machine with 4 cores, I would not recommend running MPI at all, you will be way too limited in memory. ATK will take advantage of the 4 cores automatically, in OpenMP. If you want better performance, and run properly in parallel, you should consider running on a Linux cluster, or at least a couple of Windows machines connected in a cluster (although, as mentioned, support for this is limited at this point).
-
Ok, thank you Anders Blom! The Linux cluster should be the first and best choice for HPC. :)