Hi Bruno, Thanks for the reply. It would be great if you could do me the favor. Since I'm still trying to master the library, not sure if extra parallel computation library is needed besides *hdf5 and boost* for output and toolkits.
To my knowledge, dealii image is built from source based on the official ubuntu image with a ppa repo including all dependent libraries, so it should be easy to *replace the base image to cuda*, say nvidia/cuda:11.3.1-devel-ubuntu20.04, which already have cuda support within. Wonder if it is feasible to *migrate the master image to cuda base so both GPU and CPU parallel could be integrated to one container*. The only con is that the image may be a bit too large. *Spack *is very promising for the cluster production environment and I've already trying to set it up on the HPC. Though I still prefer to docker while doing test and dev considering its convenience. Updates would be posted here if any one of them works. Regards, Chengjiang Yin Bruno Turcksin <[email protected]> 于2021年6月21日周一 下午8:41写道： > Chengjiang, > > On Saturday, June 19, 2021 at 2:38:24 PM UTC-4 [email protected] wrote: > >> >> So is there a possible way to enable the flag without rebuild and install >> the whole library or does the dealii team has any plans to release an extra >> version based on nvidia/cuda image? > > To enable cuda you need to reconfigure and rebuild deal.II. I can create a > new image with CUDA enabled but you need to tell me which packages you > need. The current docker images are based on a ppa, so I will need to > create a new image from scratch. > > >> >> It is also possible to run cuda toolkit directly without installing it >> within the container for singularity >> <https://aditishenoy.github.io/posts/blog3_nvidiacudasing/> (sorry I >> didn't find better articles) and nvidia >> <https://docs.nvidia.com/datacenter/cloud-native/driver-containers/overview.html> >> itself, which is worth trying yet not fully matured. >> > I haven't tried singularity in a while but it always _almost_ works but > then it's a pain to make it work like it should. > > >> In all, *do we have better and simpler solutions for GPU supported >> programming with dealii?* Any comments are welcome! >> > I do all my development in Docker but you could also use spack > https://github.com/dealii/dealii/wiki/deal.II-in-Spack which is great > when it works. > > Best, > > Bruno > > > -- > The deal.II project is located at http://www.dealii.org/ > For mailing list/forum options, see > https://groups.google.com/d/forum/dealii?hl=en > --- > You received this message because you are subscribed to a topic in the > Google Groups "deal.II User Group" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/dealii/WQDCw4FJ_kE/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/dealii/0dfde45d-a3d3-4864-a365-cade171953b6n%40googlegroups.com > <https://groups.google.com/d/msgid/dealii/0dfde45d-a3d3-4864-a365-cade171953b6n%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/CAGk2TyuSEg_i3iR-Pi8HztLZ_vHEp7O2o%2BTBzGenKPr5D_AZVA%40mail.gmail.com.
