Skip to content
Snippets Groups Projects

Compress hdf5 files and perform same tests with MPI.

Merged Compress hdf5 files and perform same tests with MPI.
fix_check_interactions into master
Merged James Willis requested to merge fix_check_interactions into master
Compare
and
1 file
+ 131
− 8
Compare changes
  • Side-by-side
  • Inline
examples/check_interactions.sh
+ 131
− 8
#!/bin/bash -l
#
# Runs the SedovBlast_3D and EAGLE_12 examples using naive, serial and vectorised particle interactions. Then compares the output between versions using check_ngbs.py
# Runs the SedovBlast_3D and EAGLE_12 examples using naive, serial and vectorised particle interactions.
# Then compares the output between versions using check_ngbs.py. Test is performed with and without MPI.
echo
echo "# Running SedovBlast_3D and EAGLE_12 with naive interactions and neighbour logging, 16 thread"
@@ -19,7 +20,7 @@ cd examples/SedovBlast_3D/
./getGlass.sh
python makeIC.py
../swift -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10
../swift -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv sedov_0000.hdf5 sedov_naive.hdf5
@@ -28,7 +29,7 @@ cd ../EAGLE_12/
# Link to ICs
ln -s /gpfs/data/Swift/web-storage/ICs/EAGLE_ICs_12.hdf5 EAGLE_ICs_12.hdf5
../swift -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10
../swift -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv eagle_0000.hdf5 eagle_12_naive.hdf5
@@ -44,13 +45,13 @@ make clean; make -j 6
cd examples/SedovBlast_3D/
../swift -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10
../swift -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv sedov_0000.hdf5 sedov_serial.hdf5
cd ../EAGLE_12/
../swift -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10
../swift -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv eagle_0000.hdf5 eagle_12_serial.hdf5
@@ -66,7 +67,7 @@ make clean; make -j 6
cd examples/SedovBlast_3D/
../swift -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10
../swift -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv sedov_0000.hdf5 sedov_vec.hdf5
@@ -97,7 +98,7 @@ fi
cd ../EAGLE_12/
../swift -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10
../swift -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv eagle_0000.hdf5 eagle_12_vec.hdf5
@@ -121,8 +122,130 @@ fi
if python ../check_ngbs.py eagle_12_serial.hdf5 eagle_12_vec.hdf5
then
echo "EAGLE_12 comparison between serial and vectorised passed"
exit 0
else
echo "EAGLE_12 comparison between serial and vectorised failed"
exit 1
fi
# Now run the same test using MPI
# Runs the SedovBlast_3D and EAGLE_12 examples using naive, serial and vectorised particle interactions. Then compares the output between versions using check_ngbs.py
unset I_MPI_HYDRA_BOOTSTRAP
echo
echo "# Running SedovBlast_3D and EAGLE_12 with naive interactions and neighbour logging, 16 thread, 4 MPI ranks"
echo
cd ../../
# Naive interactions run
./configure --with-metis --enable-debug-interactions=1024 --disable-vec --enable-naive-interactions
make clean; make -j 6
cd examples/SedovBlast_3D/
mpirun -np 4 ../swift_mpi -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv sedov_0000.hdf5 sedov_naive.hdf5
cd ../EAGLE_12/
mpirun -np 4 ../swift_mpi -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv eagle_0000.hdf5 eagle_12_naive.hdf5
cd ../../
echo
echo "# Running SedovBlast_3D and EAGLE_12 with serial interactions and neighbour logging, 16 thread, 4 MPI ranks"
echo
# Serial interactions run
./configure --with-metis --enable-debug-interactions=1024 --disable-vec
make clean; make -j 6
cd examples/SedovBlast_3D/
mpirun -np 4 ../swift_mpi -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv sedov_0000.hdf5 sedov_serial.hdf5
cd ../EAGLE_12/
mpirun -np 4 ../swift_mpi -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv eagle_0000.hdf5 eagle_12_serial.hdf5
cd ../../
echo
echo "# Running SedovBlast_3D and EAGLE_12 with vectorised interactions and neighbour logging, 16 thread, 4 MPI ranks"
echo
# Vectorised interactions run
./configure --with-metis --enable-debug-interactions=1024
make clean; make -j 6
cd examples/SedovBlast_3D/
mpirun -np 4 ../swift_mpi -s -t 16 -n 5 sedov.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv sedov_0000.hdf5 sedov_vec.hdf5
# Compare outputs
if python ../check_ngbs.py sedov_naive.hdf5 sedov_serial.hdf5
then
echo "SedovBlast_3D comparison between naive and serial passed (MPI)"
else
echo "SedovBlast_3D comparison between naive and serial failed (MPI)"
exit 1
fi
if python ../check_ngbs.py sedov_naive.hdf5 sedov_vec.hdf5
then
echo "SedovBlast_3D comparison between naive and vectorised passed (MPI)"
else
echo "SedovBlast_3D comparison between naive and vectorised failed (MPI)"
exit 1
fi
if python ../check_ngbs.py sedov_serial.hdf5 sedov_vec.hdf5
then
echo "SedovBlast_3D comparison between serial and vectorised passed (MPI)"
else
echo "SedovBlast_3D comparison between serial and vectorised failed (MPI)"
exit 1
fi
cd ../EAGLE_12/
mpirun -np 4 ../swift_mpi -s -t 16 -n 5 eagle_12.yml -P SPH:h_tolerance:10 -P Snapshots:compression:7
mv eagle_0000.hdf5 eagle_12_vec.hdf5
# Compare outputs
if python ../check_ngbs.py eagle_12_naive.hdf5 eagle_12_serial.hdf5
then
echo "EAGLE_12 comparison between naive and serial passed (MPI)"
else
echo "EAGLE_12 comparison between naive and serial failed (MPI)"
exit 1
fi
if python ../check_ngbs.py eagle_12_naive.hdf5 eagle_12_vec.hdf5
then
echo "EAGLE_12 comparison between naive and vectorised passed (MPI)"
else
echo "EAGLE_12 comparison between naive and vectorised failed (MPI)"
exit 1
fi
if python ../check_ngbs.py eagle_12_serial.hdf5 eagle_12_vec.hdf5
then
echo "EAGLE_12 comparison between serial and vectorised passed (MPI)"
exit 0
else
echo "EAGLE_12 comparison between serial and vectorised failed (MPI)"
exit 1
fi