From 1701e3a6e3d2b6659aacf2f35a65031b80d7963e Mon Sep 17 00:00:00 2001
From: Juan Escobar <escj@aero.obs-mip.fr>
Date: Wed, 1 Apr 2020 16:06:01 +0200
Subject: [PATCH] Juan 01/04/2020:For Irene-fr/Skylake/Knl, update configure to
use oenmpi/2.0.4
---
A-INSTALL | 14 ++++++++++----
src/configure | 6 +++---
src/job_make_examples_BullX_irene | 10 +++++-----
3 files changed, 18 insertions(+), 12 deletions(-)
diff --git a/A-INSTALL b/A-INSTALL
index 3989b7b57..b51c893bc 100644
--- a/A-INSTALL
+++ b/A-INSTALL
@@ -602,12 +602,12 @@ sbatch job_make_examples_BullX_occigen
# but with a commun disk space , connect to :
#
# - ssh irene-fr : for Intel SkyLake/KNL processors
-# On Intel processors the MPI use is INTELMPI
+# On Intel processors the MPI use is OPENMPI/2.0.4
# the configure will generate a
-# profile_mesonh-LXifort-R8I4-MNH-V5-4-3-MPIINTEL-O2
+# profile_mesonh-LXifort-R8I4-MNH-V5-4-3-MPIAUTO-O2
#
# - ssh irene-amd : for AMD , processors
-# On AMD processors the MPI use is OPENMPI
+# On AMD processors the MPI use is OPENMPI/4.02
# the configure will generate a
# profile_mesonh-LXifort-R8I4-MNH-V5-4-3-AMD-MPIAUTO-O2
#
@@ -616,8 +616,14 @@ sbatch job_make_examples_BullX_occigen
#
# - to run the test case examples run
+#
+# On intel Skylake
+ccc_msub job_make_examples_BullX_irene
+# On intel Knl
+ccc_msub -q knl job_make_examples_BullX_irene
-ccc_msub job_make_examples_BullX_irene{_AMD}
+# On intel AMD
+ccc_msub job_make_examples_BullX_irene_AMD
#
# at ECMWF on cca ( CRAY/XC30 ) :
diff --git a/src/configure b/src/configure
index 9a6a4f7a2..8b84e5045 100755
--- a/src/configure
+++ b/src/configure
@@ -123,16 +123,15 @@ module load ncl_ncarg/6.3.0
*Intel*) # Irene Intel core
export MNH_ARCH=`echo $ARCH | grep LX`
export ARCH=${MNH_ARCH:-LXifort}
- export VER_MPI=${VER_MPI:-MPIINTEL}
+ export VER_MPI=${VER_MPI:-MPIAUTO}
export OPTLEVEL=${OPTLEVEL:-O2}
export MVWORK=${MVWORK:-NO}
export VER_CDF=${VER_CDF:-CDFAUTO}
export MNHENV=${MNHENV:-"
module purge
module load intel/18.0.3.222
-module load mpi/intelmpi/2018.0.3.222
+module load mpi/openmpi/2.0.4
export SLURM_CPU_BIND=none
-export I_MPI_PIN_PROCESSOR_LIST=all:map=spread
"}
;;
*AMD*) # Irene AMD core
@@ -147,6 +146,7 @@ export I_MPI_PIN_PROCESSOR_LIST=all:map=spread
module purge
module load intel/19.0.5.281
module load mpi/openmpi/4.0.2
+export SLURM_CPU_BIND=none
# Set some openmpi variable for pb with nb of cores >> 1024
export OMPI_MCA_coll_hcoll_enable=0
export HCOLL_ENABLE_MCAST_ALL=0
diff --git a/src/job_make_examples_BullX_irene b/src/job_make_examples_BullX_irene
index b002dbb37..231bdbf7f 100755
--- a/src/job_make_examples_BullX_irene
+++ b/src/job_make_examples_BullX_irene
@@ -1,8 +1,7 @@
#!/bin/bash
#MSUB -J Examples
#MSUB -N 2 # nodes number
-#MSUB -n 4 # CPUs number (on all nodes)
-#MSUB --exclusive
+#MSUB -n 4 # CPUs number (on all nodes)
#MSUB -o Examples.eo%j #
#MSUB -e Examples.eo%j #
#MSUB -T 3600 # time limit
@@ -11,6 +10,7 @@
#MSUB -x # exclusive usage of nodes
#MSUB -X # enable X forwarding -> force 'ccc_msub -i ' to get it work
#MSUB -q skylake
+##MSUB -q knl
# Echo des commandes
ulimit -c 0
@@ -21,14 +21,14 @@ set +x
# Nom de la machine
hostname
-. ../conf/profile_mesonh-LXifort-R8I4-MNH-V5-4-3-MPIINTEL-O2
+. ../conf/profile_mesonh-LXifort-R8I4-MNH-V5-4-3-MPIAUTO-O2
set -x
module list
-export MONORUN="Exec ccc_mprun -n 1 "
-export MPIRUN="Exec ccc_mprun -n 4 "
+export MONORUN="Mpirun -tag-output --report-bindings -np 1 "
+export MPIRUN="Mpirun -tag-output --report-bindings -np 4 "
export POSTRUN="echo "
cd $SRC_MESONH/MY_RUN/KTEST/003_KW78
--
GitLab