Skip to content

YML syntax

YML syntax #2

name: Continuous integration in containers
on: [push, pull_request, workflow_dispatch]
jobs:
Containerized-CI:
runs-on: ubuntu-22.04
strategy:
fail-fast: false
matrix:
compiler: [ifort, ifx, nvfortran]
include:
# Set flags for Intel Fortran Compiler Classic
- compiler: ifort
fcflags: -m64 -g -traceback -heap-arrays -assume realloc_lhs -extend-source 132 -check bounds,uninit,pointers,stack -stand f08
# Set flags for Intel Fortran Compiler
- compiler: ifx
fcflags: -debug -traceback -O0 -heap-arrays -assume realloc_lhs -extend-source 132 -stand f08
# Set flags for NVIDIA Fortran compiler
- compiler: nvfortran
fcflags: -Mallocatable=03 -Mstandard -Mbounds -Mchkptr -Kieee -Mchkstk
# Set container images
- compiler: ifort
image: ghcr.io/earth-system-radiation/rte-rrtmgp-ci:oneapi
- compiler: ifx
image: ghcr.io/earth-system-radiation/rte-rrtmgp-ci:oneapi
- compiler: nvfortran
image: ghcr.io/earth-system-radiation/rte-rrtmgp-ci:nvhpc
container:
image: ${{ matrix.image }}
env:
F90: ${{ matrix.compiler }}
FC: ${{ matrix.compiler }}
F90FLAGS: ${{ matrix.fcflags }}
# Make variables:
NFHOME: /opt/netcdf-fortran
ATOL: 0.0
RTOL: 0.0
KGO_VERSION: v002
steps:
###############################################################################
# Build COSP and retrieve input and test files
###############################################################################
# Build COSP2 driver. Intel Fortran stores automatic arrays in the stack
# by default, whereas GNU Fortran stores them in the heap. This can cause
# segmentation faults with ifort, especially in memory-intensive applications
# like COSP. We tell ifort to use heap arrays.
- name: Build driver
run: |
source /opt/intel/oneapi/setvars.sh || true
${F90} --version
cd build
make -j driver
# Retrieve and expand large data files
- name: Retrieve data files
run: |
GDFILE='https://docs.google.com/uc?export=download&id=17eK4_DVEvFOE9Uf6siXJDpWZJKT1aqkU'
OUTPATH=driver/data/inputs/UKMO/cosp_input.um_global.nc.gz
wget --no-check-certificate $GDFILE -O $OUTPATH
gunzip ${OUTPATH}
cd driver/data/inputs/UKMO
md5sum -c cosp_input.um_global.nc.md5
cd ${GITHUB_WORKSPACE}
GDFILE='https://docs.google.com/uc?export=download&id=1s5Ha6Hqnv_hWbRUs8KQpJ4Lxy8uvJDar'
OUTPATH=driver/data/outputs/UKMO/cosp2_output.um_global.gfortran.kgo.$KGO_VERSION.nc.gz
wget --no-check-certificate $GDFILE -O $OUTPATH
gunzip ${OUTPATH}
cd driver/data/outputs/UKMO
md5sum -c cosp2_output.um_global.gfortran.kgo.$KGO_VERSION.nc.md5
cd ${GITHUB_WORKSPACE}
GDFILE='https://docs.google.com/uc?export=download&id=11dKcIL3EQr7s6jbo4f9GsoW0SufesGbq'
OUTPATH=driver/data/outputs/UKMO/cosp2_output_um.gfortran.kgo.$KGO_VERSION.nc.gz
wget --no-check-certificate $GDFILE -O $OUTPATH
gunzip ${OUTPATH}
cd driver/data/outputs/UKMO
md5sum -c cosp2_output_um.gfortran.kgo.$KGO_VERSION.nc.md5
cd ${GITHUB_WORKSPACE}
GDFILE='https://docs.google.com/uc?export=download&id=1kY1lRgzd0UhDiQef2u-VgTQql_iut3U2'
OUTPATH=driver/data/outputs/UKMO/cosp2_output.um_global_model_levels.gfortran.kgo.$KGO_VERSION.nc.gz
wget --no-check-certificate $GDFILE -O $OUTPATH
gunzip ${OUTPATH}
cd driver/data/outputs/UKMO
md5sum -c cosp2_output.um_global_model_levels.gfortran.kgo.$KGO_VERSION.nc.md5
###############################################################################
# Run COSP2 tests. We could run both tests in one step, but
# doing it this way the output is easier to interpret.
###############################################################################
# 1. Basic test
- name: Basic test, UM global snapshot
run: |
cd driver/run
./cosp2_test cosp2_input_nl.txt
./cosp2_test cosp2_input_nl.um_global.txt
###############################################################################
# Compare results against known good outputs. As above,
# we split it in as many steps as tests.
###############################################################################
# 1. Basic test
- name: Basic against known good output (KGO)
run: |
cd driver
KGO=data/outputs/UKMO/cosp2_output_um.gfortran.kgo.$KGO_VERSION.nc
TST=data/outputs/UKMO/cosp2_output_um.nc
python compare_to_kgo.py ${KGO} ${TST} --atol=${ATOL} --rtol=${RTOL}
# 2. UM global snapshot. The approach used for the basic test would needed
# large tolerances for the ifort compiler. We keep tolerances small for ifort,
# and then we test against the output table.
- name: UM global against known good output (KGO)
run: |
cd driver
KGO=data/outputs/UKMO/cosp2_output.um_global.gfortran.kgo.$KGO_VERSION.nc
TST=data/outputs/UKMO/cosp2_output.um_global.nc
ATOL=1.0e-20
RTOL=0.0006
OUTTST=data/outputs/UKMO/cosp2_output.um_global.out
OUTKGO=data/outputs/UKMO/cosp2_output.um_global.${F90}.kgo.${KGO_VERSION}.out
python compare_to_kgo.py ${KGO} ${TST} --atol=${ATOL} --rtol=${RTOL} \
--noerror=True --stats_file=${OUTTST}
diff ${OUTKGO} ${OUTTST}
###############################################################################
# Produce plots when it fails during global snapshot tests,
# and create a tarball with outputs.
###############################################################################
- name: Produce plots and create tarball
if: failure()
run: |
TST_MLEV=data/outputs/UKMO/cosp2_output.um_global_model_levels.nc
cd driver
if [[ -e data/outputs/UKMO/cosp2_output.um_global.nc ]]; then
python plot_test_outputs.py
fi
if [[ -e data/outputs/UKMO/cosp2_output.um_global_model_levels.nc ]]; then
python plot_test_outputs.py --tst_file=$TST_MLEV
fi
cd data/outputs/UKMO
tar --ignore-failed-read -czf outputs.UKMO.tgz cosp2_output.um_global.nc \
cosp2_output_um.nc cosp2_output.um_global_model_levels.nc *.png \
cosp2_output.um_global.out
ls -lh
###############################################################################
# Make output files available if any test fails
###############################################################################
- name: Upload output file if test fails
if: failure()
uses: actions/[email protected]
with:
name: outputs.UKMO.tgz
path: driver/data/outputs/UKMO/outputs.UKMO.tgz