# Fenics in Docker automatic parallelization

**URL:** <https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783>\
**Category:** Uncategorized\
**Created:** [October 24, 2019, 8:40pm UTC](https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783 "2019-10-24T20:40:00Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![amit112amit](https://avatars.discourse-cdn.com/v4/letter/a/9e8a1a/32.png) [@amit112amit](https://fenicsproject.discourse.group/u/amit112amit)\
**Post date:** [October 24, 2019, 8:40pm UTC](https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783/1 "2019-10-24T20:40:01Z")

</div>

Hi all.

I am running `fenics` in a Docker container on Arch Linux. In the past, I have used the `conda` version of fenics. I have vague memories that the `conda` version would automatically use all the cores of my laptop when I would solve a problem with large number of elements. On the other hand, the current version in Docker container uses only one core. My question is if anyone else has observed the same thing and how to make the Docker container version use all the cores? I tried changing the number of CPUs allowed to the `docker run` command but that does not change anything.

Regards,  
Amit

---

<div class="post-metadata">

**Author:** ![RossMech](https://yyz2.discourse-cdn.com/free1/user_avatar/fenicsproject.discourse.group/rossmech/32/390_2.png) [@RossMech](https://fenicsproject.discourse.group/u/RossMech)\
**Post date:** [October 25, 2019, 10:42am UTC](https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783/2 "2019-10-25T10:42:23Z")

</div>

Did you try  
mpi run -n (number of threads) python (your script) command?

---

<div class="post-metadata">

**Author:** ![bmf](https://avatars.discourse-cdn.com/v4/letter/b/df705f/32.png) [@bmf](https://fenicsproject.discourse.group/u/bmf)\
**Post date:** [October 25, 2019, 12:10pm UTC](https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783/3 "2019-10-25T12:10:39Z")

</div>

I think that this [answer](https://fenicsproject.discourse.group/t/mumps-in-parallel-slower-than-in-serial/662/2) about running code in parallel may be useful.

---

<div class="post-metadata">

**Author:** ![amit112amit](https://avatars.discourse-cdn.com/v4/letter/a/9e8a1a/32.png) [@amit112amit](https://fenicsproject.discourse.group/u/amit112amit)\
**Post date:** [October 26, 2019, 10:04pm UTC](https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783/4 "2019-10-26T22:04:18Z")

</div>

Thank you very much. I tried `mpirun`. My driver reads the mesh from a `XDMF` file that stores the mesh in a `HDF5` file. I got a bunch of errors:

```auto
HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) MPI-process 0:
  #000: ../../../src/H5F.c line 491 in H5Fcreate(): unable to create file
    major: File accessibilty
    minor: Unable to open file
  #001: ../../../src/H5Fint.c line 1168 in H5F_open(): unable to lock the file or initialize file structure
    major: File accessibilty
    minor: Unable to open file
  #002: ../../../src/H5FD.c line 1821 in H5FD_lock(): driver lock request failed
    major: Virtual File Layer
    minor: Cannot update object
  #003: ../../../src/H5FDsec2.c line 939 in H5FD_sec2_lock(): unable to flock file, errno = 11, error message = 'Resource temporarily unavailable'
    major: File accessibilty
    minor: Bad file ID accessed
HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) MPI-process 0:
  #000: ../../../src/H5F.c line 491 in H5Fcreate(): unable to create file
    major: File accessibilty
    minor: Unable to open file
  #001: ../../../src/H5Fint.c line 1168 in H5F_open(): unable to lock the file or initialize file structure
    major: File accessibilty
    minor: Unable to open file
  #002: ../../../src/H5FD.c line 1821 in H5FD_lock(): driver lock request failed
    major: Virtual File Layer
    minor: Cannot update object
  #003: ../../../src/H5FDsec2.c line 939 in H5FD_sec2_lock(): unable to flock file, errno = 11, error message = 'Resource temporarily unavailable'
    major: File accessibilty
    minor: Bad file ID accessed
HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) MPI-process 0:
  #000: ../../../src/H5F.c line 491 in H5Fcreate(): unable to create file
    major: File accessibilty
    minor: Unable to open file
  #001: ../../../src/H5Fint.c line 1168 in H5F_open(): unable to lock the file or initialize file structure
    major: File accessibilty
    minor: Unable to open file
  #002: ../../../src/H5FD.c line 1821 in H5FD_lock(): driver lock request failed
    major: Virtual File Layer
    minor: Cannot update object
  #003: ../../../src/H5FDsec2.c line 939 in H5FD_sec2_lock(): unable to flock file, errno = 11, error message = 'Resource temporarily unavailable'
    major: File accessibilty
    minor: Bad file ID accessed

```

The driver runs in spite of these messages and uses 4 threads!

---

<div class="post-metadata">

**Author:** ![amit112amit](https://avatars.discourse-cdn.com/v4/letter/a/9e8a1a/32.png) [@amit112amit](https://fenicsproject.discourse.group/u/amit112amit)\
**Post date:** [October 26, 2019, 10:05pm UTC](https://fenicsproject.discourse.group/t/fenics-in-docker-automatic-parallelization/1783/5 "2019-10-26T22:05:41Z")

</div>

Thank you for your reply. I started a Bash shell in the docker container and did  
`echo $OMP_NUM_THREADS`. I got an empty line which means the this environment variable is not set. I tried setting it to 4 and then run my code. But it was still using only one core of my laptop. But using `mpirun -n4` indeed gave me 4 threads with `OMP_NUM_THREADS=1`.