MPI Variants: OSX - Githubissues

conda-forge / openpmd-api-feedstock

A conda-smithy repository for openpmd-api.

BSD 3-Clause "New" or "Revised" License

1 stars 6 forks source link

MPI Variants: OSX #28

Closed ax3l closed 4 years ago

ax3l commented 5 years ago

Follow-up to #26: enable the three variants

nompi (default)
openmpi
mpich

for OSX as well.

Allows all Linux and OSX users to install the following:

conda install openpmd-api default, nompi variant
conda install openpmd-api=*=mpi* any mpi-compatible implementation
conda install openpmd-api=*=mpi_mpich* pick mpich variant. Same result as conda install mpich openpmd-api=*=mpi* .
conda install openpmd-api=1.13=nompi* explicitly forbid MPI. Usually not necessary, as the build number offset causes nompi to be preferred unless MPI is explicitly requested.

Checklist

[x] Used a fork of the feedstock to propose changes
[x] Bumped the build number (if the version is unchanged)
[ ] Reset the build number to 0 (if the version changed)
[x] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
[x] Ensured the license file is being packaged.

conda-forge-linter commented 5 years ago

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

ax3l commented 5 years ago

Hm, still something in the ADIOS2 libs on OSX that fails the MPI tests. Maybe https://github.com/conda-forge/adios-feedstock/pull/21 helps.

minrk commented 5 years ago

I'm not sure, but the error in the build:

Running tests...
$BUILD_PREFIX/bin/ctest --force-new-ctest-process 
Test project $SRC_DIR/build
    Start 1: Serial.Core
1/4 Test #1: Serial.Core ......................   Passed    0.06 sec
    Start 2: Serial.Auxiliary
2/4 Test #2: Serial.Auxiliary .................   Passed    0.02 sec
    Start 3: Serial.SerialIO
3/4 Test #3: Serial.SerialIO ..................***Failed    0.04 sec
Attempting to use an MPI routine before initializing MPICH

makes me think that perhaps the test is not being invoked with mpiexec. Is that possible? Or maybe the parallel tests are not compatible with --force-new-ctest-process? Those are just guesses, though.

ax3l commented 5 years ago

Yes, it looks like they are not started with mpiexec on OSX, but actually it spawns the right amount of processes. It looks more to me that those are isolated and cannot find each other, e.g. an issue with orted. Only appears on OSX, not on Linux.

ax3l commented 4 years ago

@conda-forge-admin, please rerender

ax3l commented 4 years ago

@conda-forge-admin, please rerender

conda-forge-linter commented 4 years ago

Hi! This is the friendly automated conda-forge-webservice.

I tried to re-render for you, but it looks like there was nothing to do.