This is a brand new Lenovo laptop I got as a backup because my dear Asus got broken. So I got this smaller and weaker but still cute laptop. So I wanted to make this post just to list the quickest install from 0 to a working python file.
Install python and jupyter
1. pacman -S python3
2. pacman -S jupyter-notebook
Install miniconda3 (instructions here)
3. mkdir -p ~/miniconda3
4. cd ~/miniconda3
5. wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
6. bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
7. ~/miniconda3/bin/conda init bash
Then install the packages I need. I'm pretty sure I have an exported file from this environment with the packages but this is the absolute minimum of stuff.
8.
conda create --name myenv
conda install -n myenv pip
conda install ipykernel
python3 -m ipykernel install --user --name=myenv
pacman -S gcc
pip install wheel
sudo pacman -S make
sudo pacman -S gcc-fortran
(optional conda install libgcc - no idea if this helped)
sudo pacman -S gcc-libs
pip install numpy
pip install git+https://github.com/PolyChord/PolyChordLite@master
conda install matplotlib
pip install getdist
I was getting a persistent error with libstdc++.so.6 (see here for other suggestions), here's what I did (after I installed gcc-libs and gcc-fortran)
sudo find / -name "libstdc++.so*"
conda install -c anaconda libstdcxx-ng
conda update libstdcxx-ng
ls ~/miniconda3/envs/myenv/bin/../lib
cp /usr/lib/libstdc++.so.6.0.32 ~/miniconda3/envs/myenv/bin/../lib
And then manually removing libstdc++.so.6.0.21 and libstdc++.so.6.0.29
The problem with mpi4py:
I was getting a kernel crash from inside polychord, that I know it's related to mpi4py. So I had to install it. When trying to pip install it, I woul get this error:
"The Open MPI wrapper compiler was unable to find the specified compiler
x86_64-conda_cos6-linux-gnu-cc in your PATH.
Note that this compiler was either specified at configure time or in
one of several possible environment variables.
--------------------------------------------------------------------------
failure.
removing: _configtest.c _configtest.o
error: Cannot compile MPI programs. Check your configuration!!!"
Solving it took a long time and 3 gpts!
>sudo pacman -S openmpi
>yay -S mpicc
>yay -S mpich
>echo $PATH
>conda list #look for openmpi - I had it, but obviously not working well enough
x86_64-conda_cos6-linux-gnu-cc
which gcc
/usr/bin/gcc
which mpicc
~/miniconda3/envs/BAO/bin/mpicc
export MPICC=~/miniconda3/envs/BAO/bin/mpicc
>pip install mpi4py #longer output but still and error
The Open MPI wrapper compiler was unable to find the specified compiler
x86_64-conda_cos6-linux-gnu-cc in your PATH.
Note that this compiler was either specified at configure time or in
one of several possible environment variables.
--------------------------------------------------------------------------
failure.
removing: _configtest.c _configtest.o
error: Cannot compile MPI programs. Check your configuration!!!
>which x86_64-conda_cos6-linux-gnu-cc
/i didn't have it anywhere/
>conda install x86_64-conda_cos6-linux-gnu-cc
ls ~/miniconda3/envs/BAO/bin #look for x86_64-conda_cos6-linux-gnu-cc - I have x86_64-conda-linux-gnu-cc
>export PATH=~/miniconda3/envs/BAO/bin:$PATH
>pip install mpi4py #nope, same error as above
>conda install -c conda-forge openmpi
>conda install -c conda-forge openmpi --force-reinstall
>pip install mpi4py #yay!!!
It finally worked. As you can imagine, a lot of steps, it took a while to figure it out. But now polychord is working and all is fine.
Edit (03.08.2024):
Assuming the system is already up and running with gcc, g++, mpi, mpich, openmpi, I just reinstalled the environment super quick with python 3.12
conda create --name BAOn
conda activate BAOn
conda install pip
conda install ipykernel
python3 -m ipykernel install --user --name=BAOn
which pip #to confirm you're working within the environment
pip install wheel
pip install pillow
pip install numpy==1.26 --force #to confirm it's not using the outside numpy
conda install matplotlib
pip install openmpi #note here on my computer conda install openmpi makes the calculation speed 3 times slower, so either do pip install openmpi or conda install mpich
which mpicc #to confirm the mpicc refers to the env
pip install mpi4py --force
#use force in case you have an old library
# see https://stackoverflow.com/questions/61003570/cannot-install-mpi4py-using-conda-and-specify-pre-installed-mpicc-path
pip install git+https://github.com/PolyChord/PolyChordLite@master
#this didn't work for me
#this below did work:
git clone https://github.com/PolyChord/PolyChordLite.git
cd PolyChordLite
make
pip install .
conda install jupyter
pip install emcee --force
pip install getdist
pip install anesthetic
Also, to useful commands (which I still haven't tried but I will in the future):
conda env export > BAOn.yml
conda env create -f BAOn.yml
