[QE-users] Runtime GPU memory issue of q-e-gpu- -6.4a1

[Pietro BONFA']

Dear Xiaoqin Huang,

this is likely a "out of memory" issue. It would be of help if you could:

1) run your input with verbosity = 'high', or better

2) try a (much!) smaller input such that the memory footprint reduces
from the one reported in your output (28.50 GB) to less than 10 GB. Or
even better,

3) try you input with an MPI enable version of the code on at least 4
V100s (possibly more).

As of today, pw.x only estimates the amount of *host memory* presumably
used by the simulation, but this number is clearly connected also to the
GPU memory usage.

Best regards,
Pietro Bonfà

On 9/5/19 9:55 PM, xh14 wrote:
> Hi, We have a runtime question of QE-6.4a1.
> When compiling, we used the -D options as: -D_CUDA -D__PGI -D__FFTW -D__MKL
>
> We loaded modules as GCCcore/8.3.0, CUDA/10.1.168, PGI/19.4 and
> Intel/2019a.
>
> When running a typical example, it can pass the CPU calculations, but
> meet a runtime memory issue, as the message like:
> "line 176: cudaLaunchKernel returned status 700: an illegal memory
> access was encountered".
> We do not know where the "line 176" refers to, and how to fix it.
>
> We used the slurm script, requesting 1 GPU (Volta 100), and 16 openmp
> threads on 16 CPUs.
>
> The attached file is the slurm output.
>
> Please help us of how to do next about this issue.
>
> Thanks a lot!
>
>
> Xiaoqin huang
> Rice University
> Hosuton, Texas 77030
> USA
> [text/plain]
>
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
>


--
Pietro Bonfà
Department of Mathematical, Physical and Computer Sciences,
University of Parma,
Italy

Firma il tuo 5 per mille all’Università di Parma e aiuta così i nostri studenti che vogliono realizzare un’esperienza di studio all’estero - Indica 00308780345 nella tua denuncia dei redditi.