Evaluate (and recommend) gpaw/ase from conda-forge
Evaluate the performance of the gpaw build from conda-forge against your cluster optimized build. Please post a short summary if possible (how much slower is conda-forge build for runs occupying a single node?). If the conda-forge build is fine, then recommend it as a way of installing gpaw/ase.
From https://github.com/conda-forge/gpaw-feedstock/compare/74a49879b7c27319c62dd089b1c422997f6aa5f9...c052a098fa0787ed0b76a529b875bc74c28fc226#diff-22c5a202c93f351348030b671d6e0e1fdae11f4a23eeff71551c4ef90d88bddb it seems that https://github.com/jan-janssen maintains gpaw in conda-forge so it may be possible to improve the build if needed https://github.com/conda-forge/gpaw-feedstock.
There is a talk "Introduction to Conda for (Data) Scientists Tutorial" https://www.youtube.com/watch?v=qn5zfdJtcY which shows some of conda good practices.
curl -sLO https://repo.anaconda.com/miniconda/Miniconda3-py38_4.9.2-Linux-x86_64.sh
sh Miniconda3-*-Linux-x86_64.sh -b -p $HOME/miniconda3
# Do this to avoid overwriting your default system python
echo '[ -f "$HOME/miniconda3/etc/profile.d/conda.sh" ] && source "$HOME/miniconda3/etc/profile.d/conda.sh"' >> ~/.bashrc
. ~/.bashrc
conda config --add channels conda-forge
conda config --set channel_priority strict
conda config --set env_prompt '({name}) '
conda env create -p $PWD/venv -f environment.yml
conda activate $PWD/venv
python -c "import gpaw"
mpiexec -np 2 gpaw python h2.py
conda deactivate
name: gpaw
channels:
- conda-forge
- nodefaults
dependencies:
- python=3.8
- pip=20.3
- gpaw=20.10.0
- ase=3.20.1
- numpy=1.19
- scipy=1.5
- libxc=4.3
- spglib==1.16
- netcdf4=1.5.5
- swig=4.0
- pip:
- pycodcif==3.0.1