Skip to content

SPARC ATDM Trilinos configuration and new 'cee-rhel6' env

Created by: bartlettroscoe

CC: @fryeguy52, @mhoemmen, @sebrowne, @jmgate

Description

This is a pretty significant PR. This brings in a new 'cee-rhel6' ATDM Trilinos configuration that SPARC uses to build Trilinos on the CEE LAN RHEL6 machines. This initial configuration includes a GNU, Intel, and Clang build that matches the basic GNU, Intel, and Clang post-push CI builds for SPARC.

This also adds a variation of the ATDM Trilinos configuration that I am calling the SPARC ATDM Trilinos configuration. This is done by setting the env var ATDM_ENABLE_SPARC_SETTINGS in the cee-rhel6/environment.sh file. The SPARC ATDM Trilinos configuration turns on Fortran, allows the enable the Trilinos packages ROL and ShyLU_Node, and enables the extra TPLs BinUtils, METIS, ParMETIS, CGNS, and SuperLUDist (and disables the TPL HDF5). It also changes how the Netcdf is found in a non-trivial way (SPARC tests fail if you don't enable Netcdf this way). This SPARC configuration also changes some other Trilinos configuration variables that may break the EMPIRE usage of Trilinos (but we will not know until we test this). But for the existing non-SPARC (i.e. EMPIRE) ATDM Trilinos configuration, nothing should have changed except for allowing the enable of a bunch of SEACAS subpackages and some extra STK subpackages. Hopefully allowing the enables of these extra SEACAS and STK subpackages will not cause significant new test failures or (even less likely) break the usage of the ATDM Trilinos configuration by EMPIRE.

Therefore, hopefully merging this PR should not cause many new failures in any of the existing ATDM Trilinos builds and therefore also should not negatively impact the adoption of the ATDM Trilinos configuration by EMPIRE (TRIL-222).

Motivation and Context

The ATDM Trilinos configuration needs to be upgraded to protect SPARC as well (see TRIL-212). We are just going to need to do this incrementally.

Once this PR is merged to 'develop', we can set up nightly builds of the SPARC ATDM Trilinos configuration and should start to protect the usage of Trilinos by SPARC.

How Has This Been Tested?

I ran the full set of builds for the new 'cee-rhel6' SPARC Trilinos configuration on my CEE RHEL6 machine and posted to CDash shown here. There are some new failures for ROL but nothing major.

I ran the full set of 'sems-rhel6' builds and a few of the builds on 'white' and there were no new failures.

Checklist

  • My commit messages mention the appropriate GitHub issue numbers.
  • I have updated the documentation accordingly.
  • All new and existing tests passed.

Merge request reports