I can submit multiple jobs at a time to a different program like ORCA.. Is there a way that I can do this also in Gaussian?
It is possible under Linux from the command line or from GaussView, but you have to be sure you do not engage more cores than your available cores.
In order to avoid overlaps you have to specify the number of involved cores (%nproc=...) in your input file and launch the calculations with the nohup command like this:
nice nohup g16 calc1.gjf &
and then
nice nohup g16 calc2.gjf &
With ps x you will see both calculations running. If you need to run serially two or more calculations, you can use the ; separator, as follows:
nice nohup g16 calc1_opt.gjf; nice nohup calc1_freq.gjf &
This will execute calc1_freq only when calc1_opt is over. Of course - again - you may want to execute other calculations at the same time.
I hope this helps!
Thanks for the tip ... just want to confirm one thing as I'm fairly new to this.. windows is my operating system but I did some research and found that there's a way I can run linux in windows 11... Can I still be able to follow your tip with this option? or I really need to fully change my OS to be able to do simultaneous calculations in Gaussian?
It’s true, you can run Linux on top of Windows 10/11, but obviously your system will use its resources to keep both OS running. Linux running within Windows will never be as fast as it would be running natively on the same computer. Therefore, my advice is to set up a specific machine for calculations based on a Linux OS.
Hello, Massimiliano Arca ... Is the nohup command available in all Gaussian revisions? I just want to make sure since I'm using G09 C.01... Thank you!
Yes, since the nice nohup is a linux feature and does not depends on Gaussian.
Massimiliano Arca how about linda parallel? do parallel calculations correspond to doing simultaneous calculations? are they the same? I haven't really tried Linda yet... One concern with switching to Linux is that my Gaussian is a Windows edition....
As far as I know, Linda is used for multiprocessor (cluster) parallel systems, but conceptually represents the same as addressing different cores in a multicore single processor. From the Gaussian point of view, you just specify the %nproc instruction, while it is up to Linda to send different processes to different processors.
- Badges
- Science topic
- Similar topics
- Computer Science
- Program