This repository contains a code to reproduce numerical simulations of the paper
Jan Blechta, Josef Málek, and Martin Vohralík. Localization of the W − 1, q norm for local a posteriori efficiency. Submitted, 2016. URL https://hal.inria.fr/hal-01332481.
On Linux machine with Docker installed, typing
make
will
- download Docker image https://quay.io/blechta/dolfin-tape@sha256:2425f7973b3ee5d667a1176d7a51f0f334b311298d935b5bdb2296fee7189a3a containing necessary software stack (PETSc, FEniCS, dolfin-tape); the image can be rebuilt using the files in
docker
directory, - run simulations in the Docker containers to produce log files and pdf figures,
- postprocess log files and write out
tabular.tex
.
Type make -j2
to run 2 jobs in parallel, or make -j12
to run all twelve test cases in parallel. Note that the biggest test case needs around 30GB of RAM and 30 core-hours. Completely parallel run -j12
should be possible with cca 60GB in cca 60 core-hours. Note that most of the cases can be computed in few hours and less than 8GB of memory.
Run
make INIT= RUNNER=$\(CMD\)
to avoid running in Docker container and rather use current environment. You will need
petsc4py 3.6.0
mpi4py 2.0.0
FIAT c36b6d7a988a211b04048f64c7155b0c25ed5a52
UFL 0c5b1b90498aa4f9a25fb1999463d3c1c010199a
Instant 2f355dec4142c56eb4d464e5975a1e6ea3eac493
FFC 4dc648a466ad087448a41921ade005f114e41268
DOLFIN fcf70d934d63168e5ed037678e22ac66fb2b3474
mshr 7447149c972977ff8ce2e89283d1ed0525fa2bc6
dolfin-tape 240e47f193d1f01b52f78effb29b56535274bd1b
matplotlib 1.3.1
Copyright 2015-2016, 2018 Jan Blechta
GNU LGPLv3 except the contents of docker
directory. See COPYING
, COPYING.LESSER
and docker/LICENSE
.