Skip to content

Commit 53af460

Browse files
authored
Configure IB igprof job to exit ERR when a segfault occurs instead of continuing. (#2475)
* Add environment variables and configure tensorflow without OneDNN for igprof script * Take take the tensorflow.xml from the latest MKLDNN0 release * Exit with if igprof command segfaults. Remove setup of MKLDNN0 Tensorflow * Remove scram tool info tensorflow * Remove exit $ERR
1 parent 7fe35cb commit 53af460

File tree

1 file changed

+23
-3
lines changed

1 file changed

+23
-3
lines changed

Diff for: run-ib-igprof

+23-3
Original file line numberDiff line numberDiff line change
@@ -32,7 +32,7 @@ for prof in ${PROFILES} ; do
3232
runTheMatrix.py $WF --command " -n $EVENTS --profile $prof --customise Validation/Performance/IgProfInfo.customise --no_exec" > ./runTheMatrix.log 2>&1
3333
cd $WORKFLOW*
3434
for f in $(ls *GEN_SIM.py); do
35-
igprof -mp -t cmsRunGlibC -- cmsRunGlibC $f > ${f//.py/.log} 2>&1
35+
igprof -mp -t cmsRunGlibC -- cmsRunGlibC $f > ${f//.py/.log} 2>&1 || ERR=1
3636
for g in $(ls -1 IgProf*.gz);do
3737
mv $g ${g/IgProf/igprofMEM_GC_step1}
3838
done
@@ -47,8 +47,28 @@ for prof in ${PROFILES} ; do
4747
done
4848
cd -
4949
else
50-
runTheMatrix.py $WF --command " -n $EVENTS --profile $prof" > ./runTheMatrix.log 2>&1 || ERR=1
51-
mv runTheMatrix.log $(ls -d $WORKFLOW*)
50+
export TF_ENABLE_ZENDNN_OPTS=1
51+
export OMP_NUM_THREADS=1
52+
export MALLOC_CONF=zero:true
53+
export TF_ENABLE_ONEDNN_OPTS=0
54+
55+
runTheMatrix.py $WF --command " -n $EVENTS --profile $prof --customise Validation/Performance/IgProfInfo.customise --no_exec" > ./runTheMatrix.log 2>&1
56+
cd $WORKFLOW*
57+
for f in $(ls *GEN_SIM.py); do
58+
igprof -pp -d -t cmsRun -- cmsRun $f > ${f//.py/.log} 2>&1 || ERR=1
59+
for g in $(ls -1 IgProf*.gz);do
60+
mv $g ${g/IgProf/igprofCPU_step1}
61+
done
62+
done
63+
s=1
64+
for f in $(ls -1 step*.py| sort); do
65+
igprof -pp -d -t cmsRun -- cmsRun $f > ${f//.py/.log} 2>&1 || ERR=1
66+
s=$((++s))
67+
for g in $(ls -1 IgProf*.gz);do
68+
mv $g ${g/IgProf/igprofCPU_step$s}
69+
done
70+
done
71+
cd -
5272
fi
5373
done
5474
for hpwf in $PROFILING_WORKFLOWS; do

0 commit comments

Comments
 (0)