CUDA / MPI Quantum Espresso

Compillation of Quantum Espresso with CUDA and MPI library. To get NVIDIA HPC SDK please use the link below:
developer.nvid...
For MPI library :
www.open-mpi.org/
Configure flag:
configuration in TITAN V GPU
./configure --with-cuda=$NVHPC --with-cuda-cc=70 --with-cuda-runtime=11.2 --disable-parallel --enable-openmp --with-scalapack=no --prefix=
Required modules in Supercomputer
nvhpc-nompi/21.1
For MPI you can use the following modules:
gnu9/9.4.0 and openmpi/4.1.2
In the benchmark result I have used the Intel's OneAPI and results are in Github page
github.com/rpa...
github.com/rpa...

[@mahendrashahi450]

Good initiation for the learner in the computational field of physics. Thanks sir❤

[@lalitasharma6687]

Sir what is mpi_abort was invoked on rank 0 in communicator mpi_comm_world with errorcode 1 Please help me I'm new in QE

[@LavaKafleNepal]

wow awesome

[@bhagabanswain1170]

Thank you sir for the video you have given. Sir, I have installed the QE version 7.3 in my GPU A100 80 gp ubuntu 22.04 server with cuda 12.4 and nvhpc 22.4 by follwoing your video instruction and is successful as mentioned in the video . But while running the code pw.x < pw.in > pw.out I am getting the error below please suggest.Thanking you in advance Warning: ieee_invalid is signaling Warning: ieee_divide_by_zero is signaling Warning: ieee_inexact is signaling FORTRAN STOP

[@carlosgalindo7505]

Have you checked your output file? This is a generic fortran error, and for example, I have obtained this error when I selected a pseudo that is not in my library, or I stated wrong the pseudo library locations. If you provide the exact error that is written in the output file I can help you

[@Your_ASTRO_MarY]

how is it possible..when i download qe 7.1 or qe7.2 or other versions there is no BIN folder, but you have it. Anyone can help me with this QE, please?

[@supercomputer_Nepal]

There is BIN folder in version 6.7. BIN should be there in version 7.1, 7.2 and 7.3.

[@carlosgalindo534]

hi! thank you for the video. How do you set the variable $NVHPC? If I use it the program says that my compiler is too old

[@birdinside28]

Hi, Sir, Can I run QE on Nvidia 3050ti card with 4GB VRAM? how fast will it be compared to my intel core i5 with 6 cores, 12 threads?

[@user-dq3jd2py3v]

Hello, I wanted to know if I could take a doubt. Well, I'm trying to make QE recognize my GTX 1650, but I'm not succeeding, could you help me?

[@supercomputer_Nepal]

Thank you for sharing your In general, gaming cards are not good for double precision calculations such as Quantum Espresso. GPU should first detected in OS level by driver. While compiling the QE you might have used incorrect compute capability value so it is not usable at runtime. There could be other reasons too. If you share the output log I could further help you. At first instance let me know the output of 'nvidia-smi' command without quotes.

[@user-dq3jd2py3v]

@@supercomputer_Nepal thank you for being willing to help me. I'm almost done with my degree in Physics at a university in Brazil and I'm currently learning Quantum Espresso. with this I wanted to optimize my calculations in QE through my graphic card, considering that you can help me I wanted to know where we can correspond better, so that I can show you what you ask

[@supercomputer_Nepal]

@@user-dq3jd2py3v please write me to rajendra{dot}adhikari{at}ku{dot}edu{dot}np

[@user-dq3jd2py3v]

@@supercomputer_Nepal I sent you an email with the information you requested.