Re: [QE-users] R: Data Parallelism and GPU Support for Quantum Espresso

[Prashant Govindarajan via users]

Hi! Thanks a lot for the suggestions. I tried *manypw.x*  just with 2 input
files to see if it works. Basically, my directory consists of (drive links
of text files provided)

   1. Input 1 -- espresso_0.in
   

   2. Input 2 -- espresso_1.in
   

   3. Slurm job script -- script
   

*(1 node, 8 tasks per node, 4 cpus per task, QE version 7.0)*

The command I run (on Compute Canada) is the following

srun --cpus-per-task=$SLURM_CPUS_PER_TASK manypw.x -ni 2 -i espresso


When I submit the job, Input 2 starts and it immediately stops with the
error message as described below, and Input 1 never starts.

--
 Program PWSCF v.7.0 starts on  2Jul2023 at  13:13:48

 This program is part of the open-source Quantum ESPRESSO suite
 for quantum simulation of materials; please cite
 "P. Giannozzi et al., J. Phys.:Condens. Matter 21 395502 (2009);
 "P. Giannozzi et al., J. Phys.:Condens. Matter 29 465901 (2017);
 "P. Giannozzi et al., J. Chem. Phys. 152 154105 (2020);
  URL http://www.quantum-espresso.org;,
 in publications or presentations arising from this work. More details
at
 http://www.quantum-espresso.org/quote

 Parallel version (MPI & OpenMP), running on  32 processor cores
 Number of MPI processes: 8
 Threads/MPI process: 4

 MPI processes distributed on 1 nodes
 path-images division:  nimage=   2
 R & G space division:  proc/nbgrp/npool/nimage =   4
 35167 MiB available memory on the printing compute node when the
environment starts

forrtl: Operation not permitted
forrtl: severe (28): CLOSE error, unit 6, file "Unknown"
Image  PCRoutineLineSource

manypw.x   0152002B  for__io_returnUnknown  Unknown
manypw.x   0151221F  for_close Unknown  Unknown
manypw.x   004C6BAF  Unknown   Unknown  Unknown
manypw.x   00413F67  Unknown   Unknown  Unknown
manypw.x   00413D12  Unknown   Unknown  Unknown
libc-2.30.so   2B46AD812E1B  __libc_start_main Unknown  Unknown
manypw.x   00413C2A  Unknown   Unknown  Unknown
srun: error: bc11237: task 0: Exited with exit code 28
--

I got the output file for one of the inputs as mentioned before, and it
almost reaches completion when the above error appears. The output file is
espresso_1.out
.
My understanding is that it starts for one of the inputs but when it
parallely tries to start the other input, it is trying to access something
that it does not have permissions. I'm not sure if this is a common error,
but I was not able to see a solution for this anywhere. What could be the
best way to resolve this? Please let me know if any further information is
required from my side.

Actually I have faced similar issues while using just pw.x and having a
'for loop' over multiple input crystals, so I thought I should find the
root cause of such errors.

Thanks and Regards
*Prashant Govindarajan*


On Fri, Jun 30, 2023 at 4:34 AM Paolo Giannozzi 
wrote:

> On 6/30/23 10:27, Pietro Davide Delugas wrote:
>
> > About the parallel execution: in QE, there is the manypw.x application
> > that can run many inputs in parallel.
>
> its usage is described in the header of PW/src/manypw.f90
>
> Paolo
> --
> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
> Univ. Udine, via delle Scienze 208, 33100 Udine Italy, +39-0432-558216
> ___
> The Quantum ESPRESSO community stands by the Ukrainian
> people and expresses its concerns about the devastating
> effects that the Russian military offensive has on their
> country and on the free and peaceful scientific, cultural,
> and economic cooperation amongst peoples
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users@lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
>
___
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples

[QE-users] Data Parallelism and GPU Support for Quantum Espresso

[Prashant Govindarajan via users]

Greetings,

I have been using Quantum Espresso for performing SCF calculations to
compute the band gap and energies of crystal structures. Is there a way to
perform DFT simulations on a large number of input crystals parallely using
QE, i.e., data parallelism across multiple inputs (in addition to parallel
calculations for each crystal)? For instance, what is the most optimal way
to run the pw.x command for say 5000+ crystals?

Further, I am aware that there is GPU support for QE. There are issues
while trying to install GPU-enabled QE which I am trying to figure out, but
I was wondering if it provides reasonable speedup especially when I am
dealing with multiple crystals. Also, I've been referring to
https://gitlab.com/QEF/q-e-gpu for GPU-enabled QE installation. If there
are any other useful resources please let me know.

I am using QE as an evaluation scheme for deep learning outputs, so an
effective way for GPU-based DFT simulation would be of great help.

Thanks and Regards

*Prashant Govindarajan*
*1st year PhD Student*
*Mila-Quebec AI Institute *
*Montreal, QC*
___
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users