Merge branch 'develop' of https://dci-gitlab.cines.fr/dci/abs into develop

Abinit/compile.sh

 #!/bin/bash
 
+cd /tmp/abinit-8.10.3
+make distclean
+module load intel/19.4 intelmpi/2019.4.243
 export CPATH=${MKLROOT}/include:$CPATH
 export FCFLAGS_OPENMP="-fopenmp"
-export FCFLAGS=' -extend-source -noaltparam -nofpscomp  -fopenmp'
-env CC="mpicc -fopenmp -fPIC" CXX="mpiCC -fopenmp -fPIC"  F77="mpif90 -fopenmp -fPIC" F90="mpif90 -fopenmp -fPIC" FC="mpif90 -fopenmp -fPIC" ./configure --disable-debug --enable-mpi --disable-xml --with-linalg-flavor="mkl+scalapack" --with-linalg-libs="-L${MKLROOT}/lib/intel64 -lmkl_scalapack_lp64 -lmkl_blacs_openmpi_lp64 -mkl" --with-fft-flavor=fftw3  --with-fft-libs="-lfftw3xf_intel_pic -mkl" --with-dft-flavor="libxc+atompaw+wannier90" --enable-mpi-io --with-trio-flavor="netcdf"  --enable-gw-dpc="no" --with-tardir=/tmp/tardir --prefix=$HOME/ABINIT --enable-openmp --enable-optim
+export FCFLAGS='-mtune=native -march=native -fPIC -extend-source -noaltparam -nofpscomp  -fopenmp'
+env CC="mpicc" CXX="mpicxx" F77="mpif90" F90="mpif90" FC="mpif90" ./configure --disable-debug --enable-mpi --disable-xml --with-linalg-flavor="mkl+scalapack" --with-linalg-libs="-L${MKLROOT}/lib/intel64 -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 -mkl" --with-fft-flavor=fftw3  --with-fft-libs="-lfftw3xf_intel -mkl" --with-dft-flavor="libxc+atompaw+wannier90" --enable-mpi-io --with-trio-flavor="netcdf"  --enable-gw-dpc="no" --with-tardir=/tmp/tardir --prefix=$HOME/ABINIT --enable-openmp --enable-optim
 
 make -j 40
 make install

Abinit/download.sh

@@ -2,14 +2,13 @@
 
 mkdir -p /tmp/tardir
 cd /tmp/tardir
-wget http://forge.abinit.org/fallbacks/atompaw-4.0.1.0.tar.gz
-wget http://forge.abinit.org/fallbacks/libxc-3.0.0.tar.gz
-wget http://forge.abinit.org/fallbacks/netcdf-4.1.1.tar.gz
-wget http://forge.abinit.org/fallbacks/wannier90-2.0.1.1.tar.gz
+wget -nc http://forge.abinit.org/fallbacks/atompaw-4.0.1.0.tar.gz
+wget -nc http://forge.abinit.org/fallbacks/libxc-3.0.0.tar.gz
+wget -nc http://forge.abinit.org/fallbacks/netcdf-4.1.1.tar.gz
+wget -nc http://forge.abinit.org/fallbacks/wannier90-2.0.1.1.tar.gz
 
-cd ..
-mkdir tarballs
-wget https://www.abinit.org/sites/default/files/packages/abinit-8.10.3.tar.gz
+cd /tmp
+wget -nc --no-check-certificate https://www.abinit.org/sites/default/files/packages/abinit-8.10.3.tar.gz
 tar xf abinit-8.10.3.tar.gz
 
 cd abinit-8.10.3

Abinit/machines/occigen-bdw/batch_large_ti255.slurm

@@ -14,32 +14,23 @@ module load intel/19.4 intelmpi/2019.4.243
 unset I_MPI_DAPL_DIRECT_COPY_THRESHOLD
 export MLX5_SINGLE_THREADED=0
 export MLX4_SINGLE_THREADED=0
-#module load openmpi/intel
-
-#export PATH=$SCRATCHDIR/AO2020/OPENMPI/bin:$PATH
-#export LD_LIBRARY_PATH=$SCRATCHDIR/AO2020/OPENMPI/lib:$LD_LIBRARY_PATH
-#export LIBRARY_PATH=$SCRATCHDIR/AO2020/OPENMPI/lib:$LIBRARY_PATH
-#export CPATH=$SCRATCHDIR/AO2020/OPENMPI/include:$CPATH
 
 # set max 2GB stack
 ulimit -s 2000000
 export OMP_STACKSIZE=500M
 #export KMP_DETERMINISTIC_REDUCTION=1
 
-# OCCIGEN OMP_NUM_THREADS=1 - Proc.   0 individual time (sec): cpu=      18945.1  wall=      18962.5
-# OCCIGEN OMP_NUM_THREADS=2 - Proc.   0 individual time (sec): cpu=      36255.7  wall=      18135.0
-# OCCIGEN OMP_NUM_THREADS=12 Proc.   0 individual time (sec): cpu=      40857.9  wall=       3407.8
-
 
 export KMP_LIBRARY=turnaround
 export KMP_BLOCKTIME=infinite
 export KMP_AFFINITY=verbose,granularity=fine,compact,1,0
 export OMP_NUM_THREADS=12
 export MKL_NUM_THREADS=12
-export PATH=/scratch/cnusc/dci/cirou/AO2020/ABINIT/bin:$PATH
-export LD_LIBRARY_PATH=/scratch/cnusc/dci/cirou/AO2020/ABINIT/lib:$LD_LIBRARY_PATH
 
-source /opt/software/common/intel/vtune_amplifier_2019/apsvars.sh
-srun --cpu-freq=2600000 aps abinit -i abinit_test_ti255.files > abinit_test_ti255.log.$SLURM_JOBID 2>&1
+
+#source /opt/software/common/intel/vtune_amplifier_2019/apsvars.sh
+#srun --cpu-freq=2600000 aps abinit -i abinit_test_ti255.files > abinit_test_ti255.log.$SLURM_JOBID 2>&1
+
+srun --cpu-freq=2600000 $HOME/ABINIT/bin/abinit -i abinit_test_ti255.files > abinit_test_ti255.log 2>&1
 
 #aps --report=aps_result_*

Abinit/machines/occigen-bdw/batch_small_au31.slurm

@@ -11,30 +11,21 @@ cd $SLURM_SUBMIT_DIR
 module purge
 module load intel/19.4 intelmpi/2019.4.243
 unset I_MPI_DAPL_DIRECT_COPY_THRESHOLD
-#module load openmpi/intel
-#export PATH=$SCRATCHDIR/AO2020/OPENMPI/bin:$PATH
-#export LD_LIBRARY_PATH=$SCRATCHDIR/AO2020/OPENMPI/lib:$LD_LIBRARY_PATH
-#export LIBRARY_PATH=$SCRATCHDIR/AO2020/OPENMPI/lib:$LIBRARY_PATH
-#export CPATH=$SCRATCHDIR/AO2020/OPENMPI/include:$CPATH
+export MLX5_SINGLE_THREADED=0
+export MLX4_SINGLE_THREADED=0
 
 # set max 2GB stack
 ulimit -s 2000000
-srun -l cat  /proc/sys/kernel/perf_event_paranoid 
 #export KMP_DETERMINISTIC_REDUCTION=1
-# OMP_NUM_THREADS=1 Openmpi4 Proc.   0 individual time (sec): cpu=         43.6  wall=         47.9
-# OMP_NUM_THREADS=2 Openmpi4 Proc.   0 individual time (sec): cpu=         75.1  wall=         37.8
-# OMP_NUM_THREADS=3 Openmpi4 Proc.   0 individual time (sec): cpu=        101.9  wall=         34.1
-# OMP_NUM_THREADS=4 Openmpi4 Proc.   0 individual time (sec): cpu=        130.1  wall=         32.6
-# OMP_NUM_THREADS=4 intelmpi2019 Proc.   0 individual time (sec): cpu=         67.4  wall=         17.7
 
-export MLX5_SINGLE_THREADED=0
 export KMP_LIBRARY=turnaround
 export KMP_BLOCKTIME=infinite
 export KMP_AFFINITY=granularity=fine,compact,1,0
 export OMP_NUM_THREADS=4
 export MKL_NUM_THREADS=4
-export PATH=/scratch/cnusc/dci/cirou/AO2020/ABINIT/bin:$PATH
-export LD_LIBRARY_PATH=/scratch/cnusc/dci/cirou/AO2020/ABINIT/lib:$LD_LIBRARY_PATH
+
 #source /opt/software/common/intel/vtune_amplifier_2019/apsvars.sh
 #srun --cpu-freq=2600000 aps abinit -i abinit_test_au31.files > abinit_test_au31.log.$SLURM_JOBID 2>&1
-srun --cpu-freq=2600000 abinit -i abinit_test_au31.files > abinit_test_au31.log.$SLURM_JOBID 2>&1
+srun --cpu-freq=2600000 $HOME/ABINIT/bin/abinit -i abinit_test_au31.files > abinit_test_au31.log 2>&1
+
+#aps --report=aps_result_*

Abinit/testcase_au31_small/abinit_test_au31.files

 abinit_test_au31.in
 abinit_test_au31.out
-TMP/abinit_test_au31_i
-TMP/abinit_test_au31_o
-TMP/abinit_test_au31_t
-../ATOMIC_DATA/au.hpc.paw
+abinit_test_au31_i
+abinit_test_au31_o
+abinit_test_au31_t
+au.hpc.paw
\ No newline at end of file

Abinit/testcase_au31_small/prepare.sh

+#!/bin/bash
+
+rm -f abinit_test_au31_i* abinit_test_au31_o* abinit_test_au31_t*

Abinit/testcase_au31_small/run.sh

+#!/bin/bash
+
+sbatch ../machines/occigen-bdw/batch_small_au31.slurm

Abinit/testcase_ti255_big/abinit_test_ti255.files

 abinit_test_ti255.in
 abinit_test_ti255.out
-TMP/abinit_test_ti255_i
-TMP/abinit_test_ti255_o
-TMP/abinit_test_ti255_t
-../ATOMIC_DATA/ti.hpc.paw
+abinit_test_ti255_i
+abinit_test_ti255_o
+abinit_test_ti255_t
+ti.hpc.paw
\ No newline at end of file

Abinit/testcase_ti255_big/prepare.sh

+#!/bin/bash
+
+rm -f abinit_test_ti255_i* abinit_test_ti255_o* abinit_test_ti255_t*

Abinit/testcase_ti255_big/run.sh

+#!/bin/bash
+
+sbatch ../machines/occigen-bdw/batch_large_ti255.slurm

Gromacs/README.md

@@ -15,7 +15,7 @@ Technical information:
 * Language : C/C++/CUDA
 * Parallelism : MPI + OpenMP + CUDA
 * GPU acceleration : yes  
-* Scalability : high
+* Scalability : med
 * Vectorization: high
 
 MPI-Bound Code
@@ -55,11 +55,24 @@ source machines/occigen-bdw/env
 Run and validate the simulation:
 --------------------------------
 
-For each test case, given in a separate folder (e.g. testcase_ion_channel_small and testcase_BSG_large), you can find three scripts:
-
-*  prepare.sh: prepare the simulation (move data to the right location, recompile some minor changes, ...)
-*  run.sh : run the application and print out the evaluated metric in a file called results (first numeric column representing ns/day corresponds to the evaluated metric to take into account)
-*  validate.sh: validation of the simulation for (if the evaluated metric is print out and the simulation is completed with 20 000 steps for testcase_ion_channel_small and 1 200 000 steps for testcase_BSG_large, the bench is considered validated)
+For each test case, given in a separate folder (e.g. testcase_ion_channel_small, testcase_ion_channel_large and testcase_BSG_large), you can find three scripts:
+
+1.  prepare.sh: prepare the simulation (move data to the right location, recompile some minor changes, ...)
+2.  run.sh : run the application and print out the evaluated metric in a file called results (first numeric column representing ns/day corresponds to the evaluated metric to take into account)
+3.  validate.sh: validation of the simulation. 
+
+*  The bench testcase_ion_channel_small is considered validated if:
+    * the evaluated metric is print out ;
+    * the simulation is completed with  20 000 steps ;
+    * the files small_ion_channel.log, small_ion_channel.edr, results and slurm*.out are present in the directory results-gromacs-ion_channel_small.
+*  The bench testcase_ion_channel_large is considered validated if:
+    *  the evaluated metric is print out ;
+    *  the simulation is completed with 1 000 000 steps ;
+    *  the files large_ion_channel.log, large_ion_channel.edr, results and slurm*.out are present in the directory results-gromacs-ion_channel_large.
+*  The bench testcase_BSG_large is considered validated if:
+    *  the evaluated metric is print out ;
+    *  the simulation is completed with 1 200 000 steps ;
+    *  the files benchBSG.cpt, benchBSG.trr, benchBSG.xtc, benchBSG.log, benchBSG.edr, results and slurm*.out are present in the directory results-gromacs-BSG_large.
 
 For running and validating the simulation, one should be able to do:
 ```

Gromacs/machines/occigen-bdw/batch_large_BSG.slurm

-d #!/bin/bash
+#!/bin/bash
 #SBATCH -J benchBSG
 #SBATCH -C BDW28
 #SBATCH -t 01:30:00

Gromacs/testcase_BSG_large/validate.sh

@@ -12,6 +12,12 @@ cd $run_gromacs_dir_name
 perf=`grep Performance: results | awk '{ print $2 }'`
 end_sim=`grep "Statistics over 1200001 steps using 12001 frames" benchBSG.log`
 
+mkdir results-gromacs-BSG_large
+cp benchBSG.* results-gromacs-BSG_large/.
+cp slurm* results-gromacs-BSG_large/.
+cp results results-gromacs-BSG_large/.
+cd results-gromacs-BSG_large
+
 if [ -z "$perf" ] || [ -z "$end_sim" ]
 then
     echo "bench is not validated"

Gromacs/testcase_ion_channel_large/validate.sh

@@ -12,6 +12,12 @@ cd $run_gromacs_dir_name
 perf=`grep Performance results | awk '{ print $2 }'`
 end_sim=`grep "Statistics over 1000001 steps using 100001 frames" large_ion_channel.log`
 
+mkdir results-gromacs-ion_channel_large
+cp large_ion_channel* results-gromacs-ion_channel_large/.
+cp slurm* results-gromacs-ion_channel_large/.
+cp results results-gromacs-ion_channel_large/.
+cd results-gromacs-ion_channel_large
+
 if [ -z "$perf" ] || [ -z "$end_sim" ]
 then
     echo "bench is not validated"

Gromacs/testcase_ion_channel_small/validate.sh

@@ -12,6 +12,12 @@ cd $run_gromacs_dir_name
 perf=`grep Performance results | awk '{ print $2 }'`
 end_sim=`grep "Statistics over 20001 steps using 2001 frames" small_ion_channel.log`
 
+mkdir results-gromacs-ion_channel_small
+cp small_ion_channel* results-gromacs-ion_channel_small/.
+cp slurm* results-gromacs-ion_channel_small/.
+cp results results-gromacs-ion_channel_small/.
+cd results-gromacs-ion_channel_small
+
 if [ -z "$perf" ] || [ -z "$end_sim" ]
 then
     echo "bench is not validated"