X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/2a81e806265c18e1c1f8ff624a37f317105b3904..4339cfd5b7bd278326b9e7c6467dd72cb87ea456:/src/smpi/smpirun.in diff --git a/src/smpi/smpirun.in b/src/smpi/smpirun.in index 5b2b05350a..c558d38d68 100755 --- a/src/smpi/smpirun.in +++ b/src/smpi/smpirun.in @@ -1,6 +1,6 @@ #!/usr/bin/env sh -# Copyright (c) 2007-2018. The SimGrid Team. All rights reserved. +# Copyright (c) 2007-2020. 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. @@ -31,21 +31,25 @@ SIMOPTS="--cfg=surf/precision:1e-9 --cfg=network/model:SMPI" usage () { cat < -hostfile program [program-options] + $0 [OPTIONS] -platform -hostfile -replay [program] [program-options] Options: -keep-temps # don't remove the generated files after execution -wrapper # use command to run the program (e.g. "valgrind" or "gdb --args") + -gdb # run within GDB (-wrapper "gdb --args" -keep-temps) + -lldb # run within LLDB (-wrapper "lldb --" -keep-temps) + -vgdb # run within Valgrind+GDB (-wrapper "valgrind --vgdb=yes --vgdb-error=0" -keep-temps) -map # display the machine on which each process rank is mapped -np # use that amount of processes from the hostfile. # By default, all processes of the hostfile are used. -no-privatize # Disable the globals privatization, that is activated by default - -trace-ti # activate time independant tracing (for replay, default in smpi_simgrid.txt) + -trace-ti # activate time independent tracing (for replay, default in smpi_simgrid.txt) -trace # activate tracing (Paje, default in smpi_simgrid.trace) -trace-comment # put a comment on the top of the trace file -trace-comment-file # put file contents on the top of the trace file as comment -trace-grouped # group MPI processes by location -trace-resource # trace resource utilization -trace-file # name of the tracefile (simgrid_smpi.trace) - -ext # additional parameter (reserved) + -replay # replays a trace instead of actually executing an application -version # Displays the SimGrid version (human readable) -git-version # Displays the git hash of SimGrid @@ -63,39 +67,39 @@ then exit fi -EXTOPT="" WRAPPER="" HOSTFILE="" HOSTFILETMP=0 MAPOPT=0 REPLAY=0 +QUIET="" unset pid trapped_signals="HUP INT QUIT ILL ABRT SEGV FPE ALRM TERM USR1 USR2 BUS" die () { - printf '[%s] ** error: %s. Aborting.\n' "$(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 "${APPLICATIONTMP}" APPLICATIONTMP="" fi fi @@ -107,14 +111,14 @@ smpirun_trap() { # Cleanup and kill the child process: smpirun_cleanup - if ! [ -z "$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 $$ + trap - "$trapped_signals" + kill -"$sig" $$ # This should never happen: kill -ABRT $$ @@ -122,7 +126,7 @@ smpirun_trap() { } for s in $trapped_signals; do - trap "smpirun_trap $s" $s + trap 'smpirun_trap $s' "$s" done while true; do @@ -153,8 +157,8 @@ while true; do fi shift 2 ;; - "-ext") - EXTOPT="$2" + "-replay") + APP_TRACES="$2" shift 2 ;; "-no-privatize") @@ -195,14 +199,36 @@ while true; do shift 1 ;; "-keep-temps") - KEEP="true" - SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes" + KEEP="true" + SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes" shift 1 ;; + "-quiet") + QUIET="true" + shift 1 + ;; "-wrapper") WRAPPER="$2" shift 2 ;; + "-gdb") + WRAPPER="gdb --args" + KEEP="true" + SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes" + shift 1 + ;; + "-vgdb") + WRAPPER="valgrind --vgdb=yes --vgdb-error=0" + KEEP="true" + SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes" + shift 1 + ;; + "-lldb") + WRAPPER="lldb --" + KEEP="true" + SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes" + shift 1 + ;; "-help" | "--help" | "-h") usage exit 0 @@ -232,18 +258,27 @@ while true; do esac done -# check if we still have at least one parameter beyond options -if [ $# -eq 0 ] -then - echo "Error: no program to execute!" - usage - exit -fi +if [ -n "${APP_TRACES}" ] ; then + if [ $# -eq 0 ] ; then + EXEC="@SMPIREPLAYMAIN@" + else + EXEC="$1" + shift + fi +else + # check if we still have at least one parameter beyond options + if [ $# -eq 0 ] + then + echo "Error: no program to execute!" + usage + exit + fi -EXEC="$1" -shift + EXEC="$1" + shift +fi -# steel --cfg and --logs options +# steal --cfg and --logs options while [ $# -gt 0 ]; do case "$1" in "--cfg="*|"--log="*) @@ -269,42 +304,57 @@ fi if [ -z "${HOSTFILE}" ] ; then HOSTFILETMP=1 HOSTFILE="$(mktemp smpitmp-hostfXXXXXX)" - perl -ne 'print "$1\n" if /.*.*/' ${PLATFORM} > ${HOSTFILE} - # put all > ${HOSTFILE} + @PYTHON_EXECUTABLE@ -c ' +import xml.etree.ElementTree as ET +import sys +import re + +tree = ET.parse(sys.stdin) + +for elem in tree.findall(".//host"): + print(elem.attrib["id"]) + +for elem in tree.findall(".//cluster"): + prefix = elem.attrib["prefix"] + radical = elem.attrib["radical"] + suffix = elem.attrib["suffix"] + for r in radical.split(","): + m = re.match("^([^-]*?)-([^-]*)$", r) + if m: + for i in range(int(m.group(1)), int(m.group(2))): + print(prefix + str(i) + suffix) + else: + print(prefix + r + suffix) + ' < "${PLATFORM}" > "${HOSTFILE}" fi UNROLLEDHOSTFILETMP=0 -#parse if our lines are terminated by :num_process -multiple_processes=$(grep -c ":" $HOSTFILE) -if [ "${multiple_processes}" -gt 0 ] ; then +# parse if our lines are terminated by :num_process +if grep -q ':' "$HOSTFILE" ; then UNROLLEDHOSTFILETMP=1 UNROLLEDHOSTFILE="$(mktemp smpitmp-hostfXXXXXX)" - perl -ne ' do{ for ( 1 .. $2 ) { print "$1\n" } } if /(.*?):(\d+).*/' ${HOSTFILE} > ${UNROLLEDHOSTFILE} + @PYTHON_EXECUTABLE@ -c ' +import sys +import re + +for line in sys.stdin: + m = re.match("(.*):(.*)", line) + if m: + for i in range(0, int(m.group(2))): + print(m.group(1)) + else: + print(line.strip()) +' < "${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 '${HOSTFILE}' is empty" fi @@ -313,7 +363,7 @@ if [ -z "${NUMPROCS}" ] ; then 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 ${hostfile_procs} processes in your hostfile..." >&2 fi @@ -321,7 +371,7 @@ fi if [ -z "${PLATFORM}" ]; then PLATFORMTMP="$(mktemp smpitmp-platfXXXXXX)" - cat > ${PLATFORMTMP} < "${PLATFORMTMP}" < @@ -329,28 +379,30 @@ if [ -z "${PLATFORM}" ]; then PLATFORMHEAD i=${NUMPROCS} - while [ $i -gt 0 ]; do - echo " " >> ${PLATFORMTMP} - echo " " >> ${PLATFORMTMP} - echo " " >> ${PLATFORMTMP} + while [ "$i" -gt 0 ]; do + { + echo " " + echo " " + echo " " + } >> "${PLATFORMTMP}" i=$((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 " " >> ${PLATFORMTMP} + while [ "$j" -gt 0 ]; do + if [ "$i" -eq "$j" ]; then + echo " " >> "${PLATFORMTMP}" else - echo " " >> ${PLATFORMTMP} + echo " " >> "${PLATFORMTMP}" fi j=$((j - 1)) done i=$((i - 1)) done - cat >> ${PLATFORMTMP} <> "${PLATFORMTMP}" < PLATFORMFOOT @@ -363,22 +415,21 @@ fi APPLICATIONTMP="$(mktemp smpitmp-appXXXXXX)" #APPLICATIONTMP="app.xml" -cat > ${APPLICATIONTMP} < "${APPLICATIONTMP}" < APPLICATIONHEAD ##---- cache hostnames of hostfile--------------- -if [ -n "${HOSTFILE}" ] && [ -f ${HOSTFILE} ]; then - hostnames=$(cat ${HOSTFILE} | tr '\n\r' ' ') +if [ -n "${HOSTFILE}" ] && [ -f "${HOSTFILE}" ]; then + hostnames=$(< "${HOSTFILE}" tr '\n\r' ' ') fi -if [ "${EXTOPT}" = "smpi_replay" ]; then - APP_TRACES=$PROC_ARGS - if [ -n "${APP_TRACES}" ] && [ -f "${APP_TRACES}" ]; then - hosttraces=$(cat ${APP_TRACES} | tr '\n\r' ' ' ) - NUMTRACES=$(cat ${APP_TRACES} | wc -l) +if [ -n "${APP_TRACES}" ]; then + if [ -f "${APP_TRACES}" ]; then + hosttraces=$(< "${APP_TRACES}" tr '\n\r' ' ' ) + NUMTRACES=$(< "${APP_TRACES}" wc -l) REPLAY=1 else printf "File not found: %s\n" "${APP_TRACES:-\${APP_TRACES\}}" >&2 @@ -398,7 +449,7 @@ if [ -n "${HAVE_SEQ}" ]; then SEQ=$(${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 @@ -425,20 +476,21 @@ do echo " - " >> ${APPLICATIONTMP} + " >> "${APPLICATIONTMP}" if [ ${REPLAY} = 1 ]; then - if [ ${NUMTRACES} -gt 1 ]; then - echo " " >> ${APPLICATIONTMP} + echo " " >> "${APPLICATIONTMP}" + if [ "${NUMTRACES}" -gt 1 ]; then + echo " " >> "${APPLICATIONTMP}" else - echo " " >> ${APPLICATIONTMP} + echo " " >> "${APPLICATIONTMP}" fi else - echo ${XML_ARGS} >> ${APPLICATIONTMP} + echo "${XML_ARGS}" >> "${APPLICATIONTMP}" fi - echo " " >> ${APPLICATIONTMP} + echo " " >> "${APPLICATIONTMP}" done -cat >> ${APPLICATIONTMP} <> "${APPLICATIONTMP}" < APPLICATIONFOOT ##-------------------------------- end DEFAULT APPLICATION -------------------------------------- @@ -475,9 +527,10 @@ if [ -n "${TRACE_ACTIVE}" ]; then fi ##---------------------- end SMPI TRACING OPTIONS --------------------------------- +# Do not remove, this variable may be used by user code (e.g. StarPU) export SMPI_GLOBAL_SIZE=${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 ${HOSTFILE} kept." fi @@ -495,9 +548,9 @@ fi # * The job is launched in the background in order to be able to handle signals. # # * The FD 3 is used to temporarily store FD 1. This is because the shell connects FD 1 to /dev/null when the command -# is launched in the background: this can be overriden in bash but not in standard bourne shell. +# is launched in the background: this can be overridden in bash but not in standard bourne shell. exec 3<&0 -${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 @@ -513,17 +566,17 @@ 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