CI: Switch intel build to ifx #183
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #183 +/- ##
=======================================
Coverage 92.83% 92.83%
=======================================
Files 47 47
Lines 6785 6785
Branches 762 762
=======================================
Hits 6299 6299
Misses 474 474
Partials 12 12
Flags with carried forward coverage won't be shown. Click here to find out more. |
danielhollas
force-pushed
the
ci-setup-fortran
branch
from
2299df0 to
c7748c5
Compare
danielhollas
changed the title
| @@ -88,7 +89,6 @@ jobs: | |||
| path: ~/pfunit/build/installed | |||
| # To force a pFUnit rebuild (bust the cache), make a change to install_pfunit.sh | |||
| key: ${{ runner.os }}-ubuntu20.04-pfunit-gfortran${{ env.GCC_V }}-${{ hashFiles('dev_scripts/install_pfunit.sh') }} | |||
| save-always: true | |||
There was a problem hiding this comment.
This option is deprecated (and is likely not working as intented anyway)
| runs-on: ubuntu-22.04 | ||
| needs: basic_build | ||
| strategy: | ||
| fail-fast: false | ||
| matrix: | ||
| gcc_v: [9, 11, 12] |
There was a problem hiding this comment.
Let's have one less build. The optimized builds never caught any issues. At some point I'll switch to just having a one of these.
| @@ -0,0 +1 @@ | |||
| 12 | |||
There was a problem hiding this comment.
There are some numerical differences in the ifx build in the restart files for this test, so we decrease required numerical accuracy to 1e-12. The NUM_THRE file is read in the tests/numdiff.py script.
| @@ -27,7 +27,7 @@ namespace { | |||
|
|
|||
| // two body parameters | |||
| const double reoh = 0.958649; | |||
| const double thetae = 104.3475; | |||
| // const double thetae = 104.3475; | |||
There was a problem hiding this comment.
Got a compiler warning that this param is not used.
| sudo apt-get install ${{ env.APT_PACKAGES }} | ||
| source /opt/intel/oneapi/setvars.sh | ||
| printenv >> $GITHUB_ENV | ||
| - uses: fortran-lang/setup-fortran@v1 |
There was a problem hiding this comment.
For now I haven't switched the gfortran builds to this GH action since it does not support gfortran v7.
@JanosJiri Are some Prague clusters still on gcc-7?
There was a problem hiding this comment.
Prague cluster status:
Krypton: gcc version 10.2.1 20210110 (Debian 10.2.1-6)
Neon: gcc version 4.7.2 (Debian 4.7.2-5)
Argon: gcc version 4.4.5 (Debian 4.4.5-8)
Argon is almost dead, Neon is still used but Polách...
| sudo apt-get install ${{ env.APT_PACKAGES }} | ||
| source /opt/intel/oneapi/setvars.sh | ||
| printenv >> $GITHUB_ENV | ||
| - uses: fortran-lang/setup-fortran@v1 |
There was a problem hiding this comment.
Prague cluster status:
Krypton: gcc version 10.2.1 20210110 (Debian 10.2.1-6)
Neon: gcc version 4.7.2 (Debian 4.7.2-5)
Argon: gcc version 4.4.5 (Debian 4.4.5-8)
Argon is almost dead, Neon is still used but Polách...
Partially addresses #176
Instead of testing with the old intel compiled (
ifort) let's test with the new LLVM-basedifx.