Is Quantum Espresso already installed on this system? Your link includes a general submission script for SGE, which will queue up the job for you. Was their something specific you were having trouble understanding? — Tyberius ♦ 4 hours ago

@Tyberius It's been a long time since I used SGE but I could easily give instructions for a beginner to run their first calculation on LoadLeveler, SLURM, PBS, etc. Perhaps we could generalize this a bit and have a thread for beginners to learn how to submit their first job with each type of scheduler, in one-topic-per-answer fashion? — Nike Dattani 3 hours ago

@Tyberius I don't think Quantum Espresso is already installed on the system. I actually can not find the list of software available at the moment. But I will ask them. What I am looking for is something like what are the generic steps in order to run your input file and all. Honestly I am a bit confused too but Anoop has done a fantastic job, but they are using SGE. And yes a thread discussing how to run jobs for different schedulers seems like something people like me will find really helpful. Thanks — Parmeet Singh EP 066 2 hours ago

You may have the command module spider available, which should list all loadable programs and their version(s). — Tyberius ♦ 2 hours ago

Another command that would help if you don't have module spider, is module avail. However I'm starting to think we might be better of helping you in this chat room. — Nike Dattani 2 hours ago

I actually knew about those commands but I am sort of having problem with logining into the system(using ssh), I think there is something wrong with the password. So I will try that after this. But if you want I could start using the chat box instead. @NikeDattani — Parmeet Singh EP 066 9 mins ago

Usually when we get to about 7 comments on a post, there's the risk that the SE system will try to move us to Chat. Right now, both the question and Anoop's answer have 7+ comments, so it would be a good idea for you to write your next response in the chat room! — Nike Dattani 1 min ago

Your answer is really great but sadly the facility I have access to uses SGE for there schedule manager. But I wanna ask you two questions: 1. How do you know which node you are running your calculation on. Because you are not supposed to run any jobs on the login node. Like do you choose them by name? 2. Is quantum espresso parallelised during the compilation and we do not have to worry about it during each run? — Parmeet Singh EP 066 4 hours ago

@ParmeetSinghEP066 Hi! yes you are right you shouldnt run calculations on the login node. when you submit your job the HPC will automatically allocate a node for the same. In order to know the node you are running on you can check the queue of runnig calculations usually they would have the number of node there. QE is compiled in parallel mode if you have mpirun installed and if you are using a parallell intel compiler (Im not entirely sure about the latter). A sample command to compile QE in parallell is to download it. run ./configure CC=icc F90=ifort MPIF90=mpiifort within the QE folder — Anoop A Nair 4 hours ago

@ParmeetSinghEP066 Then you can use make pw to compile pwSCF modules or make all to make everything. If its successful you'd have executables inside the bin folder. — Anoop A Nair 4 hours ago

@ParmeetSinghEP066 You will not know which node something is running on until you submit Anoop's script and the scheduler process it and assigns your job to a node. You can find out which node it's running on, after it starts running, by running squeue -u parmeet if your username is parmeet. You are allowed to run Anoop's submission script (the one he gave you in this answer) from the login node, and what that will do, is send the job to the compute nodes. You're imply not supposed to run pw.x on the login node, you have to run sbatch submission.sh instead. QE is already parallelized. — Nike Dattani 4 hours ago

Ok thanks. I will run the jobs using scheduler and that should not do any problem. I was actually a bit worried because I read somewhere that if you run your calculation on the login node it can bring it down somehow. — Parmeet Singh EP 066 2 hours ago

@ParmeetSinghEP066 dont worry about bringing the entire HPC down :) . Your jobs submitted to the login node usually gets cancelled and in the worst case scenario ((after multiple attempts to submit to the login node)) your account maybe be suspended with an email inquiry email from the system admin. — Anoop A Nair 2 hours ago

@AnoopANair are you using the same HPC system as Parmeet? The limits for jobs on login nodes will differ from system to system. In some cases, submitting jobs to the login node can be a serious problem for the system. However if Parmeet runs a submission script to a scheduler, rather than just running the pw.x command directly from the command line, I think he should be okay. — Nike Dattani 2 hours ago

@NikeDattani Since the HPC parmeet is using, provides access to the users at the university, its highly unlikely that measures to prevent actions of one user to bring down the entire HPC are not in place. The HPC at my university instantly cancels all mutli-core jobs submitted to the login node. I was just explaining the several possibilities that may exist but (ParmeetSinghEP066) its always nice to ask the sys admin before experimenting with it :) — Anoop A Nair 22 mins ago

I agree, though I once caused a computing cluster at University of Waterloo to crash, and a lot of jobs ended prematurely because of it. I had told the sys admin in advance to limit my account so that this wouldn't happen, and he told me he did that successfully, yet something didn't work and I caused a professor's calculations to get interrupted. So just because the HPC system is run by a university doesn't mean it's always safe to run jobs on the login nodes. I agree they should put in measures to prevent anything bad from happening, but the users also have to be responsible :) — Nike Dattani 2 mins ago