#!/usr/bin/env sh
-# Copyright (c) 2007-2019. The SimGrid Team. All rights reserved.
+# Copyright (c) 2007-2021. The SimGrid Team. All rights reserved.
# This program is free software; you can redistribute it and/or modify it
# under the terms of the license (GNU LGPL) which comes with this package.
DEFAULT_LOOPBACK_LATENCY="0.000004s"
DEFAULT_NETWORK_BANDWIDTH="$((26 * 1024 * 1024))Bps"
DEFAULT_NETWORK_LATENCY="0.000005s"
-DEFAULT_NUMPROCS="4"
DEFAULT_SPEED="100flops"
LOOPBACK_BANDWIDTH="${DEFAULT_LOOPBACK_BANDWIDTH}"
SIMOPTS="--cfg=surf/precision:1e-9 --cfg=network/model:SMPI"
+SMPITMPDIR="$(dirname $(mktemp -u))"
+
#usage to print the way this script should be called
usage () {
cat <<EOF
-np <numprocs> # use that amount of processes from the hostfile.
# By default, all processes of the hostfile are used.
-no-privatize # Disable the globals privatization, that is activated by default
- -trace-ti # activate time independant tracing (for replay, default in smpi_simgrid.txt)
+ -tmpdir # Directory used to store temporary files. Defaults to system's.
+ -trace-ti # activate time independent tracing (for replay, default in smpi_simgrid.txt)
-trace # activate tracing (Paje, default in smpi_simgrid.trace)
-trace-comment <comment> # put a comment on the top of the trace file
-trace-comment-file <file> # put file contents on the top of the trace file as comment
HOSTFILETMP=0
MAPOPT=0
REPLAY=0
+QUIET=""
unset pid
trapped_signals="HUP INT QUIT ILL ABRT SEGV FPE ALRM TERM USR1 USR2 BUS"
die () {
- printf '[%s] ** error: %s. Aborting.\n' "$(basename $0)" "$*" >&2
+ printf '[%s] ** error: %s. Aborting.\n' "$(basename "$0")" "$*" >&2
exit 1
}
smpirun_cleanup()
{
if [ -z "${KEEP}" ] ; then
- if [ -z "${PLATFORM}" -a -n "$PLATFORMTMP" ]; then
- rm -f ${PLATFORMTMP}
+ if [ -z "${PLATFORM}" ] && [ -n "$PLATFORMTMP" ]; then
+ rm -f "${PLATFORMTMP}"
PLATFORMTMP=""
fi
- if [ ${HOSTFILETMP} = 1 -a -n "$HOSTFILE" ] ; then
- rm -f ${HOSTFILE}
+ if [ ${HOSTFILETMP} = 1 ] && [ -n "$HOSTFILE" ] ; then
+ rm -f "${HOSTFILE}"
HOSTFILE=""
fi
- if [ ${UNROLLEDHOSTFILETMP} = 1 -a -n "$UNROLLEDHOSTFILE" ] ; then
- rm -f ${UNROLLEDHOSTFILE}
+ if [ "${UNROLLEDHOSTFILETMP}" = 1 ] && [ -n "$UNROLLEDHOSTFILE" ] ; then
+ rm -f "${UNROLLEDHOSTFILE}"
UNROLLEDHOSTFILE=""
fi
- if [ -n ${APPLICATIONTMP} ]; then
- rm -f ${APPLICATIONTMP}
+ if [ -n "${APPLICATIONTMP}" ]; then
+ rm -f "${APPLICATIONTMP}"
APPLICATIONTMP=""
fi
fi
# Cleanup and kill the child process:
smpirun_cleanup
- if ! [ -z "$pid" ]; then
- kill -TERM $pid
+ if [ -n "$pid" ]; then
+ kill -TERM "$pid"
fi
unset pid
# Raise the same signal again (remove the traps first):
trap - $trapped_signals
- kill -$sig $$
+ kill -"$sig" $$
# This should never happen:
kill -ABRT $$
}
for s in $trapped_signals; do
- trap "smpirun_trap $s" $s
+ trap "smpirun_trap $s" "$s"
done
while true; do
MAPOPT=1
shift 1
;;
+ "-tmpdir")
+ SMPITMPDIR="$2"
+ shift 1
+ ;;
"-trace")
TRACE_ACTIVE="true"
shift 1
SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes"
shift 1
;;
+ "-quiet")
+ QUIET="true"
+ shift 1
+ ;;
"-wrapper")
WRAPPER="$2"
shift 2
esac
done
+#setup tmp dir
+SIMOPTS="$SIMOPTS --cfg=smpi/tmpdir:$SMPITMPDIR"
+export LD_LIBRARY_PATH="$SMPITMPDIR:$LD_LIBRARY_PATH"
+
if [ -n "${APP_TRACES}" ] ; then
if [ $# -eq 0 ] ; then
EXEC="@SMPIREPLAYMAIN@"
shift
fi
-# steel --cfg and --logs options
+# steal --cfg and --logs options
while [ $# -gt 0 ]; do
case "$1" in
"--cfg="*|"--log="*)
print(prefix + str(i) + suffix)
else:
print(prefix + r + suffix)
- ' < ${PLATFORM} > ${HOSTFILE}
+ ' < "${PLATFORM}" > "${HOSTFILE}"
fi
UNROLLEDHOSTFILETMP=0
# parse if our lines are terminated by :num_process
-if grep -q ':' $HOSTFILE ; then
+if grep -q ':' "$HOSTFILE" ; then
UNROLLEDHOSTFILETMP=1
UNROLLEDHOSTFILE="$(mktemp smpitmp-hostfXXXXXX)"
@PYTHON_EXECUTABLE@ -c '
print(m.group(1))
else:
print(line.strip())
-' < ${HOSTFILE} > ${UNROLLEDHOSTFILE}
+' < "${HOSTFILE}" > "${UNROLLEDHOSTFILE}"
if [ ${HOSTFILETMP} = 1 ] ; then
- rm ${HOSTFILE}
+ rm "${HOSTFILE}"
HOSTFILETMP=0
fi
HOSTFILE=$UNROLLEDHOSTFILE
fi
# Don't use wc -l to compute it to avoid issues with trailing \n at EOF
-hostfile_procs=$(grep -c "[a-zA-Z0-9]" $HOSTFILE)
-if [ ${hostfile_procs} = 0 ] ; then
+hostfile_procs=$(grep -c "[a-zA-Z0-9]" "$HOSTFILE")
+if [ "${hostfile_procs}" = 0 ] ; then
die "the hostfile '${HOSTFILE}' is empty"
fi
NUMPROCS=$hostfile_procs
fi
-if [ ${NUMPROCS} -gt ${hostfile_procs} ] ; then
+if [ "${NUMPROCS}" -gt "${hostfile_procs}" ] ; then
echo "You requested to use ${NUMPROCS} ranks, but there is only ${hostfile_procs} processes in your hostfile..." >&2
fi
if [ -z "${PLATFORM}" ]; then
PLATFORMTMP="$(mktemp smpitmp-platfXXXXXX)"
- cat >
${PLATFORMTMP} <<PLATFORMHEAD
+ cat >
"${PLATFORMTMP}" <<PLATFORMHEAD
<?xml version='1.0'?>
<!DOCTYPE platform SYSTEM "https://simgrid.org/simgrid.dtd">
<platform version="4.1">
PLATFORMHEAD
i=${NUMPROCS}
- while [ $i -gt 0 ]; do
- echo " <host id=\"host$i\" speed=\"${SPEED}\"/>" >> ${PLATFORMTMP}
- echo " <link id=\"loop$i\" bandwidth=\"${LOOPBACK_BANDWIDTH}\" latency=\"${LOOPBACK_LATENCY}\"/>" >> ${PLATFORMTMP}
- echo " <link id=\"link$i\" bandwidth=\"${NETWORK_BANDWIDTH}\" latency=\"${NETWORK_LATENCY}\"/>" >> ${PLATFORMTMP}
+ while [ "$i" -gt 0 ]; do
+ {
+ echo " <host id=\"host$i\" speed=\"${SPEED}\"/>"
+ echo " <link id=\"loop$i\" bandwidth=\"${LOOPBACK_BANDWIDTH}\" latency=\"${LOOPBACK_LATENCY}\"/>"
+ echo " <link id=\"link$i\" bandwidth=\"${NETWORK_BANDWIDTH}\" latency=\"${NETWORK_LATENCY}\"/>"
+ } >> "${PLATFORMTMP}"
i=$((i - 1))
done
i=${NUMPROCS}
- while [ $i -gt 0 ]; do
+ while [ "$i" -gt 0 ]; do
j=${NUMPROCS}
- while [ $j -gt 0 ]; do
- if [ $i -eq $j ]; then
- echo " <route src=\"host$i\" dst=\"host$j\"><link_ctn id=\"loop$i\"/></route>" >> ${PLATFORMTMP}
+ while [ "$j" -gt 0 ]; do
+ if [ "$i" -eq "$j" ]; then
+ echo " <route src=\"host$i\" dst=\"host$j\"><link_ctn id=\"loop$i\"/></route>" >> "${PLATFORMTMP}"
else
- echo " <route src=\"host$i\" dst=\"host$j\"><link_ctn id=\"link$i\"/><link_ctn id=\"link$j\"/></route>" >> ${PLATFORMTMP}
+ echo " <route src=\"host$i\" dst=\"host$j\"><link_ctn id=\"link$i\"/><link_ctn id=\"link$j\"/></route>" >> "${PLATFORMTMP}"
fi
j=$((j - 1))
done
i=$((i - 1))
done
- cat >>
${PLATFORMTMP} <<PLATFORMFOOT
+ cat >>
"${PLATFORMTMP}" <<PLATFORMFOOT
</zone>
</platform>
PLATFORMFOOT
APPLICATIONTMP="$(mktemp smpitmp-appXXXXXX)"
#APPLICATIONTMP="app.xml"
-cat >
${APPLICATIONTMP} <<APPLICATIONHEAD
+cat >
"${APPLICATIONTMP}" <<APPLICATIONHEAD
<?xml version='1.0'?>
<!DOCTYPE platform SYSTEM "https://simgrid.org/simgrid.dtd">
<platform version="4.1">
APPLICATIONHEAD
##---- cache hostnames of hostfile---------------
-if [ -n "${HOSTFILE}" ] && [ -f ${HOSTFILE} ]; then
- hostnames=$(cat ${HOSTFILE} | tr '\n\r' ' ')
+if [ -n "${HOSTFILE}" ] && [ -f "${HOSTFILE}" ]; then
+ hostnames=$(< "${HOSTFILE}" tr '\n\r' ' ')
fi
if [ -n "${APP_TRACES}" ]; then
if [ -f "${APP_TRACES}" ]; then
- hosttraces=$(cat ${APP_TRACES} | tr '\n\r' ' ' )
- NUMTRACES=$(cat ${APP_TRACES} | wc -l)
+ hosttraces=$(< "${APP_TRACES}" tr '\n\r' ' ' )
+ NUMTRACES=$(< "${APP_TRACES}" wc -l)
REPLAY=1
else
printf "File not found: %s\n" "${APP_TRACES:-\${APP_TRACES\}}" >&2
SEQ=$(${HAVE_SEQ} 0 $(( NUMPROCS - 1)))
else
cnt=0
- while [ $cnt -lt ${NUMPROCS} ] ; do
+ while [ $cnt -lt "${NUMPROCS}" ] ; do
SEQ="$SEQ $cnt"
cnt=$((cnt + 1));
done
if [ ${MAPOPT} = 1 ]; then
echo "[rank $i] -> $host"
fi
-
+ {
echo " <actor host=\"${host}\" function=\"$i\"> <!-- function name used only for logging -->
<prop id=\"instance_id\" value=\"smpirun\"/>
- <prop id=\"rank\" value=\"$i\"/>" >> ${APPLICATIONTMP}
+ <prop id=\"rank\" value=\"$i\"/>"
if [ ${REPLAY} = 1 ]; then
- echo " <prop id=\"smpi_replay\" value=\"true\"/>" >> ${APPLICATIONTMP}
- if [ ${NUMTRACES} -gt 1 ]; then
- echo " <argument value=\"$(echo $hosttraces|cut -d' ' -f$j)\"/>" >> ${APPLICATIONTMP}
+ echo " <prop id=\"smpi_replay\" value=\"true\"/>"
+ if [ "${NUMTRACES}" -gt 1 ]; then
+ echo " <argument value=\"$(echo "$hosttraces"|cut -d' ' -f$j)\"/>"
else
- echo " <argument value=\"$(echo $hosttraces|cut -d' ' -f1)\"/>" >> ${APPLICATIONTMP}
+ echo " <argument value=\"$(echo "$hosttraces"|cut -d' ' -f1)\"/>"
fi
else
- echo ${XML_ARGS} >> ${APPLICATIONTMP}
+ echo "${XML_ARGS}"
fi
- echo " </actor>" >> ${APPLICATIONTMP}
+ echo " </actor>"
+ } >> "${APPLICATIONTMP}"
done
-cat >>
${APPLICATIONTMP} <<APPLICATIONFOOT
+cat >>
"${APPLICATIONTMP}" <<APPLICATIONFOOT
</platform>
APPLICATIONFOOT
##-------------------------------- end DEFAULT APPLICATION --------------------------------------
# Do not remove, this variable may be used by user code (e.g. StarPU)
export SMPI_GLOBAL_SIZE=${NUMPROCS}
-if [ -n "${KEEP}" ] ; then
- echo ${EXEC} ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP}
+if [ -n "${KEEP}" ] && [ -z "${QUIET}" ] ; then
+ echo "${EXEC}" ${PRIVATIZE} "${TRACEOPTIONS}" "${SIMOPTS}" "${PLATFORMTMP}" "${APPLICATIONTMP}"
if [ ${HOSTFILETMP} = 1 ] ; then
echo "Generated hostfile ${HOSTFILE} kept."
fi
# * The job is launched in the background in order to be able to handle signals.
#
# * The FD 3 is used to temporarily store FD 1. This is because the shell connects FD 1 to /dev/null when the command
-# is launched in the background: this can be overriden in bash but not in standard bourne shell.
+# is launched in the background: this can be overridden in bash but not in standard bourne shell.
exec 3<&0
-${WRAPPER} "@SMPIMAIN@" ${EXEC} ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP} <&3 3>&- &
+${WRAPPER} "@SMPIMAIN@" "${EXEC}" ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} "${PLATFORMTMP}" "${APPLICATIONTMP}" <&3 3>&- &
pid=$!
exec 3>&-
wait $pid
# Keep temporary files on failures to help debugging
#
if [ ${status} -ne 0 ] ; then
- if [ -z "${KEEP}" ]; then
- echo ${EXEC} ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP}
+ if [ -z "${KEEP}" ] && [ -z "${QUIET}" ]; then
+ echo "${EXEC}" ${PRIVATIZE} "${TRACEOPTIONS}" "${SIMOPTS}" "${PLATFORMTMP}" "${APPLICATIONTMP}"
if [ ${HOSTFILETMP} = 1 ] ; then
echo "Generated hostfile ${HOSTFILE} kept."
fi
if [ ${UNROLLEDHOSTFILETMP} = 1 ] ; then
echo "Generated unrolled hostfile ${UNROLLEDHOSTFILE} kept."
fi
+ KEEP=true
fi
echo "Execution failed with code ${status}."
- KEEP=true
fi
smpirun_cleanup
