WORK IN PROGRESS

NOT USE-ABLE YET

SIPG_PMMA_demo_2022

SIPG FEM simulation and comparison of linear elastic and linear viscoelastic response for PMMA used (circa 2022) as described below. This code and the docker image (see below) can be used to reproduce the data in the paper:

A Priori Analysis of a Symmetric Interior Penalty Discontinuous Galerkin Finite Element Method for a Dynamic Linear Viscoelasticity Model

By Yongseok Jang and Simon Shaw

Submitted to: Computational Methods in Applied Mathematics

The DOI and full reference will appear here if it is accepted.

Git usage

Some useful Git tips for using these codes, and a trail of how the Git and docker resources were created.

# first set up the bare (README.md, LICENSE and .gitignore) repo on Git Hub
# clone it and then add files
cd ~/to/parent/directory
git clone git@github.com:variationalform/SIPG_PMMA_demo_2022.git

This will get you the basic codes needed to run the simulation. If you want to get up and running more quickly then the docker version described below will suit your needs better.

The basic commands are something like this:

cd ./runtime/le && ./longrun_le.sh | tee ./longrun_le_out.txt
cd ./runtime/ve && ./longrun_ve.sh | tee ./longrun_ve_out.txt

These will change to a working directory for linear elasticity or viscoelasticity (le or ve) and then execute the bash scripts. The output is tee'd off to stdout as well as the *.txt text files. This could take many days to run - edit the bash scripts to just get the coarser results if you want output more quickly.

Git: general usage (some memory joggers)

git status                       # shows what has changed
git diff                         # shows differences before staging
git add ./README.md .gitignore   # stage
git commit -m 'useful message'   # commit
git diff --staged                # differences after staging
git push                         # push to remote

docker - runtime

The following sequence of commands can be used to generate a suite of results for linear elasticity and viscoelasticity using the discrete time SIPG FEM as described in the paper referenced above. These results increase mesh and time-step precision with each set taking progressively longer to run.

# log in to server, cd to where the shared directories will be, for example
cd fenics_docker/shared
docker pull variationalform/fem:SIPG_PMMA_demo_2022
# you might need this to run screen
sudo chmod o+rw /dev/pts/0
screen -S pmma # at least three: one for le, one for ve and one to monitor progress

# in screen 2
newgrp docker # you may not need this
cd fenics_docker/shared
docker run -ti --name le_pmma -v "$PWD":/home/fenics/shared \
        -w /home/fenics/local/runtime/le variationalform/fem:SIPG_PMMA_demo_2022
# in the container
./longrun_le.sh | tee ./longrun_le_out.txt
# docker rm (e.g.) 3e17 can be used for a fresh start with the container
# periodically, to get interim results, use CNTRL-Z...
cd .. && tar cvf le.tar le && mv le.tar /home/fenics/shared/ && cd le
# and then use fg to resume 

# in screen 3
newgrp docker # you may not need this
cd fenics_docker/shared
docker run -ti --name ve_pmma -v "$PWD":/home/fenics/shared \
        -w /home/fenics/local/runtime/ve variationalform/fem:SIPG_PMMA_demo_2022
# in the container
./longrun_ve.sh | tee ./longrun_ve_out.txt
# docker rm (e.g.) 3e17 can be used for a fresh start with the container
# periodically, to get interim results, use CNTRL-Z...
cd .. && tar cvf ve.tar ve && mv ve.tar /home/fenics/shared/ && cd ve
# and then use fg to resume 

These runs could take up to ten days (or even longer) depending on your hardware. The *.sh scripts can be edited to get shorter (but less accurate) runs.

You can detach from the screen session while the runs execute. For docker the commands below are useful. First, if CNTRL-D was used to exit the container then you can re-attach like this:

# list containers - note the ID on the left (e.g. adf...)
docker ps -a
# to start and then attach to a stopped container
docker start adf
docker attach adf

On the other hand if CNTRL-p CNTRL-q was used to leave the container then

docker attach adf

will get you back in it. If you CNTRL-D out of a container and then try to docker run again you'll get an error because trhe container name is already in use. For example,

docker run -ti --name le_pmma -v "$PWD":/home/fenics/runtime_le \
    -w /home/fenics/local/runtime/le variationalform/fem:SIPG_PMMA_demo_2022
docker: Error response from daemon: Conflict. The container name "/le_pmma" is
already in use by container
"7aef802a84aca5b063b9e10be9ebc17bc5165a62cc0a111f8e47ccd37381dc36".
You have to remove (or rename) that container to be able to reuse that name.
See 'docker run --help'.

In this case, note the ID and just remove the container for a fresh start:

docker rm 7a

docker - set up

For interest, this is how the image was set up. We used the following sequence of commands. The FEniCS image was selected from those available at https://quay.io/repository/fenicsproject/stable?tab=tags. It's quite old but we've been using it for years and know it to be reliable.

docker pull quay.io/fenicsproject/stable:2017.2.0.r4
# to run it when working in the shared directory (not a good idea in general)
docker run -ti --name SIPG_PMMA_demo_2022 -v "$PWD":/home/fenics/shared \
        -w /home/fenics/shared quay.io/fenicsproject/stable:2017.2.0.r4
# we'll need 'screen'
sudo apt-get update
sudo apt-get install screen
screen -S <name>
# this failed with the following message:
# Cannot open your terminal '/dev/pts/0' - please check.
# solve it with this:
sudo chmod o+rw /dev/pts/0

# screen helpers:
screen -ls && screen -r <name>
#toggle previous two C-A C-A; next terminal C-AN; switch to shell n C-A n
#C-a (shift to u/c) A <term_name>
#C-AD to detach and exit
#screen -d -r (detaches and reattaches when necessary)
#CA ? gives a list of commands

# to determine the FENICS version:
python
>>> import dolfin as df; print (df.__version__); exit()

# to set up the local files ready for running the solves...
cd ~/local
# we copy the directory structure from a remote through the share
cp -r ~/shared/runtime/ .
# it's from a mac so get rid of these things...
rm ./runtime/.DS_Store ./runtime/ve/.DS_Store ./runtime/le/.DS_Store 
ls -R

# alter RUNPATH in the sh files if necessary and then these are the run commands
fenics@329e2fdd6283:~/local/runtime/le$ ./longrun_le.sh | tee ./longrun_le_out.txt
fenics@329e2fdd6283:~/local/runtime/ve$ ./longrun_ve.sh | tee ./longrun_ve_out.txt

This has given us a working version of FEniCS in a docker image. Its runtime can be realized with a docker container. The instructions for doing that are above.

For interest,

# to run with a share but working in a local directory
docker run -ti --name SIPG_PMMA_demo_2022 -v "$PWD":/home/fenics/shared \
        -w /home/fenics/local quay.io/fenicsproject/stable:2017.2.0.r4
# in FEniCS
cp -r ../shared/runtime/ .
# you might need this to run screen
sudo chmod o+rw /dev/pts/0
screen -S <name>
# can re-attach with screen -d -r <name> (-d if not cleanly detached)
# run bash in several screens sessions, and then in each of two sessions
cd ~/local/runtime/le ; ./longrun_le.sh | tee ./longrun_le_out.txt
cd ~/local/runtime/ve ; ./longrun_ve.sh | tee ./longrun_ve_out.txt

docker hub

The tag was created on https://hub.docker.com/repository/docker/variationalform/fem with the following:

# get the tag
docker ps --filter "status=exited"
# commit it
docker commit 329
# gives
sha256:afcfbfc0646d40a89f300da5d6e4e453be5099a9875940a54835444497b59fda
# login, create and push
docker login
docker tag afcfb variationalform/fem:SIPG_PMMA_demo_2022
docker push variationalform/fem:SIPG_PMMA_demo_2022

Now a downloader can execute and run as explained at the top of this section

Alternatively, once the tar file, made like this,

docker save afcfb > SIPG_PMMA_demo_2022_docker_afcfb.tar

is available, the container can be instyanced with this:

docker load < SIPG_PMMA_demo_2022_docker_afcfb.tar
docker run -ti afcfb

TO DO: update https://hub.docker.com/repository/docker/variationalform/fem with the paper's DOI

TO DO: make the tarfile available (figshare?)