CDO

We are trying to parallelize a CDO operation as below. Even though we can get some parallelism with OpenMP, we want to see if launching multiple CDO tasks in parallel as below can provide better speedup. But we are seeing an odd behavior where even though we see the multiple CDO tasks launched, they are all running on a single core (monitoring the cpu utilization using htop). Is there any reason that launching these in parallel is not using the multiple cores on the system?

We've also tried the GNU parallel approach mentioned in the forum - https://code.mpimet.mpg.de/boards/53/topics/6672 , but still see similar behavior (i.e. using only 1 core in the system)

COUNTPP=0
for FILE in ${DATAFILELIST}
do
(
  FILEOUT=$(basename ${FILE} .nc)
  ${CDO} -s -P ${OMP_THREADS} ${GCM_REMAP},${INIDIR}/triangular-grid.nc -selname,u,v,w,pres,temp,qv,qc,qi,z_ifc${ICON_INPUT_OPTIONAL}  ${FILE} ${OUTDIR}/${YYYY_MM}/${FILEOUT}_lbc.nc
)&
    (( COUNTPP=COUNTPP+1 ))
    if [ ${COUNTPP} -ge ${MAXCORES} ]
    then
      COUNTPP=0
      wait
    fi
done
wait

Any advice would be very helpful.