From 1341964ec4592ab178b282c6fca5655e7bec39aa Mon Sep 17 00:00:00 2001
From: Richard Top <richard.topouchian@uib.no>
Date: Fri, 14 Jun 2024 09:29:17 +0000
Subject: [PATCH 1/2] Add mem per node setting in ReFrame configuration
 template

---
 reframe_config_bot.py.tmpl | 5 +++++
 test_suite.sh              | 9 ++++++++-
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/reframe_config_bot.py.tmpl b/reframe_config_bot.py.tmpl
index 0cc3e9f530..607373767a 100644
--- a/reframe_config_bot.py.tmpl
+++ b/reframe_config_bot.py.tmpl
@@ -34,6 +34,11 @@ site_configuration = {
                             'options': ['--mem={size}'],
                         }
                     ],
+                    'extras': {
+                        # Make sure to round down, otherwise a job might ask for more mem than is available
+                        # per node
+                        'mem_per_node': __MEM_PER_NODE__,
+                    },
                     'max_jobs': 1
                     }
                 ]
diff --git a/test_suite.sh b/test_suite.sh
index 2f304dd9bc..104f09fc2c 100755
--- a/test_suite.sh
+++ b/test_suite.sh
@@ -135,7 +135,7 @@ export RFM_PREFIX=$PWD/reframe_runs
 echo "Configured reframe with the following environment variables:"
 env | grep "RFM_"
 
-# Inject correct CPU properties into the ReFrame config file
+# Inject correct CPU/memory properties into the ReFrame config file
 cpuinfo=$(lscpu)
 if [[ "${cpuinfo}" =~ CPU\(s\):[^0-9]*([0-9]+) ]]; then
     cpu_count=${BASH_REMATCH[1]}
@@ -165,6 +165,13 @@ if [[ "${cpuinfo}" =~ (Core\(s\) per socket:[^0-9]*([0-9]+)) ]]; then
 else
     fatal_error "Failed to get the number of cores per socket for the current test hardware with lscpu."
 fi
+cgroup_mem_bytes=$(cat /sys/fs/cgroup/memory/slurm/uid_${UID}/job_${SLURM_JOB_ID}/memory.limit_in_bytes)
+if [[ $? -eq 0 ]]; then
+    # Convert to MiB
+    cgroup_mem_mib=$((cgroup_mem_bytes/(1024*1024)))
+else
+    fatal_error "Failed to get the memory limit in bytes from the current cgroup"
+fi
 cp ${RFM_CONFIG_FILE_TEMPLATE} ${RFM_CONFIG_FILES}
 sed -i "s/__NUM_CPUS__/${cpu_count}/g" $RFM_CONFIG_FILES
 sed -i "s/__NUM_SOCKETS__/${socket_count}/g" $RFM_CONFIG_FILES

From e05c9e70ef7a99c28b4f2c101751a24a27bf499c Mon Sep 17 00:00:00 2001
From: Richard Top <richard.topouchian@uib.no>
Date: Fri, 14 Jun 2024 09:31:28 +0000
Subject: [PATCH 2/2] Add mem per node setting in ReFrame configuration
 template

---
 test_suite.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/test_suite.sh b/test_suite.sh
index 104f09fc2c..af2c5bfb2f 100755
--- a/test_suite.sh
+++ b/test_suite.sh
@@ -177,6 +177,7 @@ sed -i "s/__NUM_CPUS__/${cpu_count}/g" $RFM_CONFIG_FILES
 sed -i "s/__NUM_SOCKETS__/${socket_count}/g" $RFM_CONFIG_FILES
 sed -i "s/__NUM_CPUS_PER_CORE__/${threads_per_core}/g" $RFM_CONFIG_FILES
 sed -i "s/__NUM_CPUS_PER_SOCKET__/${cores_per_socket}/g" $RFM_CONFIG_FILES
+sed -i "s/__MEM_PER_NODE__/${cgroup_mem_mib}/g" $RFM_CONFIG_FILES
 
 # Workaround for https://github.com/EESSI/software-layer/pull/467#issuecomment-1973341966
 export PSM3_DEVICES='self,shm'  # this is enough, since we only run single node for now