iomkl toolchain with GCC-5.4.0-2.26 and OpenMPI 1.10.3#3772
iomkl toolchain with GCC-5.4.0-2.26 and OpenMPI 1.10.3#3772boegel merged 2 commits intoeasybuilders:developfrom
Conversation
|
@OleHolmNielsen can you please also include an HPL easyconfig using this toolchain as a test case? |
|
I have created and pushed the file HPL test module for this toolchain. |
|
@OleHolmNielsen submitting of test reports is done via I'll look into test reports for this, so we can merge it in for EasyBuild v3.0. |
|
Test report by @boegel |
|
lgtm |
|
Test report by @boegel |
|
Going in, thanks @OleHolmNielsen! |
These files are based upon the {iomkl,iompi,imkl}-2016.09-GCC-4.9.3-2.25.eb files.
GCC compiler updated to GCC-5.4.0-2.26 because we already have this from the foss2016b toolchain.
OpenMPI 1.10.3 is used because we also have this in the foss2016b toolchain.
The iomkl toolchain MUST be built in a certain order:
Commands used and tested:
eb icc-2016.3.210-GCC-5.4.0-2.26.eb iccifort-2016.3.210-GCC-5.4.0-2.26.eb ifort-2016.3.210-GCC-5.4.0-2.26.eb -r
eb OpenMPI-1.10.3-Slurm-iccifort-2016.3.210-GCC-5.4.0-2.26.eb -r # For Slurm support
eb iompi-2016.09-GCC-5.4.0-2.26.eb imkl-11.3.3.210-iompi-2016.09-GCC-5.4.0-2.26.eb -r
eb iomkl-2016.09-GCC-5.4.0-2.26.eb -r
When the modules are ready, the iomkl toolchain can be loaded. The mpirun command is working correctly with one of our production codes.
Warning: If users load the iompi module, this picks up the mpirun command from the ifort compiler, and this fails miserably for reasons that we don't understand.