3 # Copyright (c) 2007-2016, The SimGrid Team. All rights reserved.
5 # This program is free software; you can redistribute it and/or modify it
6 # under the terms of the license (GNU LGPL) which comes with this package.
10 SIMGRID_VERSION="@SIMGRID_VERSION_STRING@"
11 SIMGRID_GITHASH="@SIMGRID_GITHASH@"
13 DEFAULT_LOOPBACK_BANDWIDTH="498000000Bps"
14 DEFAULT_LOOPBACK_LATENCY="0.000004s"
15 DEFAULT_NETWORK_BANDWIDTH="$((26 * 1024 * 1024))Bps"
16 DEFAULT_NETWORK_LATENCY="0.000005s"
18 DEFAULT_SPEED="100flops"
20 LOOPBACK_BANDWIDTH="${DEFAULT_LOOPBACK_BANDWIDTH}"
21 LOOPBACK_LATENCY="${DEFAULT_LOOPBACK_LATENCY}"
22 NETWORK_BANDWIDTH="${DEFAULT_NETWORK_BANDWIDTH}"
23 NETWORK_LATENCY="${DEFAULT_NETWORK_LATENCY}"
24 SPEED="${DEFAULT_SPEED}"
26 PRIVATIZE="--cfg=smpi/privatization:@HAVE_PRIVATIZATION@"
28 SIMOPTS="--cfg=surf/precision:1e-9 --cfg=network/model:SMPI"
30 #usage to print the way this script should be called
33 Usage: $0 [OPTIONS] -platform <xmldesc> -hostfile <hostfile> program [program-options]
35 -keep-temps # don't remove the generated files after execution
36 -wrapper <command> # use command to run the program (e.g. "valgrind" or "gdb --args")
37 -map # display the machine on which each process rank is mapped
38 -np <numprocs> # use that amount of processes from the hostfile.
39 # By default, all processes of the hostfile are used.
40 -no-privatize # Disable the globals privatization, that is activated by default
41 -trace-ti # activate time independant tracing (for replay, default in smpi_simgrid.txt)
42 -trace # activate tracing (Paje, default in smpi_simgrid.trace)
43 -trace-comment <comment> # put a comment on the top of the trace file
44 -trace-comment-file <file> # put file contents on the top of the trace file as comment
45 -trace-grouped # group MPI processes by location
46 -trace-resource # trace resource utilization
47 -trace-viva # generate configuration for Viva's GraphView
48 -trace-file <tracefile> # name of the tracefile (simgrid_smpi.trace)
49 -ext <value> # additional parameter (reserved)
51 -version # Displays the SimGrid version (human readable)
52 -git-version # Displays the git hash of SimGrid
54 or (deprecated usage):
55 $0 [-keep-temps] [-np <numprocs>] [-bandwidth <bytes/sec>] [-latency <secs>] program [program-options]
60 #check if we have at least one parameter
76 trapped_signals="HUP INT QUIT ILL ABRT SEGV FPE ALRM TERM USR1 USR2 BUS"
79 printf '[%s] ** error: %s. Aborting.\n' "$(basename $0)" "$*" >&2
85 if [ -z "${KEEP}" ] ; then
86 if [ -z "${PLATFORM}" -a -n "$PLATFORMTMP" ]; then
90 if [ ${HOSTFILETMP} = 1 -a -n "$HOSTFILE" ] ; then
94 if [ ${UNROLLEDHOSTFILETMP} = 1 -a -n "$UNROLLEDHOSTFILE" ] ; then
95 rm -f ${UNROLLEDHOSTFILE}
98 if [ -n ${APPLICATIONTMP} ]; then
99 rm -f ${APPLICATIONTMP}
109 # Cleanup and kill the child process:
111 if ! [ -z "$pid" ]; then
116 # Raise the same signal again (remove the traps first):
117 trap - $trapped_signals
120 # This should never happen:
125 for s in $trapped_signals; do
126 trap "smpirun_trap $s" $s
136 NETWORK_BANDWIDTH="$2"
145 if [ ! -f "${PLATFORM}" ]; then
146 die "the file '${PLATFORM}' does not exist"
152 if [ ! -f "${HOSTFILE}" ]; then
153 die "the file '${HOSTFILE}' does not exist"
159 if [ ! -f "${HOSTFILE}" ]; then
160 die "the file '${HOSTFILE}' does not exist"
182 TRACE_TI_ACTIVE="true"
189 "-trace-comment-file")
190 TRACE_COMMENT_FILE="$2"
202 TRACE_RESOURCE="true"
211 SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes"
218 "-help" | "--help" | "-h")
222 "-version" | "--version" | "-v")
223 printf '%b\n' "$SIMGRID_VERSION"
226 "-git-version" | "--git-version")
227 printf '%b\n' "$SIMGRID_GITHASH"
233 SIMOPTS="$SIMOPTS ${1%%=*}=$OPT"
238 # Nothing to do, compatibility.
247 # check if we still have at least one parameter beyond options
250 echo "Error: no program to execute!"
258 # steel --cfg and --logs options
259 while [ $# -gt 0 ]; do
264 SIMOPTS="$SIMOPTS ${1%%=*}=$OPT"
269 PROC_ARGS="${PROC_ARGS:+$PROC_ARGS }$1"
275 if [ -z "${HOSTFILE}" ] && [ -z "${PLATFORM}" ] ; then
276 echo "No hostfile nor platform specified."
281 if [ -z "${HOSTFILE}" ] ; then
283 HOSTFILE="$(mktemp smpitmp-hostfXXXXXX)"
284 perl -ne 'print "$1\n" if /.*<host.*?id="(.*?)".*?\/>.*/' ${PLATFORM} > ${HOSTFILE}
285 perl -ne 'if (/.*<cluster.*?prefix="(.*?)".*?radical="(.*?)".*?suffix="(.*?)".*/) {
286 my ($pre,$rad,$post)=($1,$2,$3);
287 for my $elm (split(",",$rad)) {
288 if ($elm=~/^([^-]*?)-([^-]*)$/) {
289 for (my $i=$1; $i<=$2;$i++) {
290 print "$pre$i$post\n";
293 print "$pre$elm$post\n";
296 } elsif (/<cluster/) {
297 die ("Unparsable cluster tag. smpirun uses a primitive regular expression to parse cluster tags. Either provide a hostfile yourself or give the attributes prefix, radical and suffix IN THAT ORDER and ON THE SAME LINE as the opening tag (<cluster)");
298 }' ${PLATFORM} >> ${HOSTFILE}
300 UNROLLEDHOSTFILETMP=0
302 #parse if our lines are terminated by :num_process
303 multiple_processes=$(grep -c ":" $HOSTFILE)
304 if [ "${multiple_processes}" -gt 0 ] ; then
305 UNROLLEDHOSTFILETMP=1
306 UNROLLEDHOSTFILE="$(mktemp smpitmp-hostfXXXXXX)"
307 perl -ne ' do{ for ( 1 .. $2 ) { print "$1\n" } } if /(.*?):(\d+).*/' ${HOSTFILE} > ${UNROLLEDHOSTFILE}
308 if [ ${HOSTFILETMP} = 1 ] ; then
312 HOSTFILE=$UNROLLEDHOSTFILE
315 # Don't use wc -l to compute it to avoid issues with trailing \n at EOF
316 hostfile_procs=$(grep -c "[a-zA-Z0-9]" $HOSTFILE)
317 if [ ${hostfile_procs} = 0 ] ; then
318 die "the hostfile '${HOSTFILE}' is empty"
321 if [ -z "${NUMPROCS}" ] ; then
322 # Use the amount of processes in the hostfile as default value for the -np parameter
323 NUMPROCS=$hostfile_procs
326 if [ ${NUMPROCS} -gt ${hostfile_procs} ] ; then
327 echo "You requested to use ${NUMPROCS} ranks, but there is only ${hostfile_procs} processes in your hostfile..." >&2
330 ##-------------------------------- DEFAULT or SPECIFIED PLATFORM --------------------------------------
331 if [ -z "${PLATFORM}" ]; then
332 PLATFORMTMP="$(mktemp smpitmp-platfXXXXXX)"
334 cat > ${PLATFORMTMP} <<PLATFORMHEAD
335 <?xml version='1.0'?>
336 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd">
337 <platform version="4.1">
338 <zone id="AS0" routing="Full">
342 while [ $i -gt 0 ]; do
343 echo " <host id=\"host$i\" speed=\"${SPEED}\"/>" >> ${PLATFORMTMP}
344 echo " <link id=\"loop$i\" bandwidth=\"${LOOPBACK_BANDWIDTH}\" latency=\"${LOOPBACK_LATENCY}\"/>" >> ${PLATFORMTMP}
345 echo " <link id=\"link$i\" bandwidth=\"${NETWORK_BANDWIDTH}\" latency=\"${NETWORK_LATENCY}\"/>" >> ${PLATFORMTMP}
350 while [ $i -gt 0 ]; do
352 while [ $j -gt 0 ]; do
353 if [ $i -eq $j ]; then
354 echo " <route src=\"host$i\" dst=\"host$j\"><link_ctn id=\"loop$i\"/></route>" >> ${PLATFORMTMP}
356 echo " <route src=\"host$i\" dst=\"host$j\"><link_ctn id=\"link$i\"/><link_ctn id=\"link$j\"/></route>" >> ${PLATFORMTMP}
363 cat >> ${PLATFORMTMP} <<PLATFORMFOOT
369 PLATFORMTMP=${PLATFORM}
371 ##-------------------------------- end DEFAULT or SPECIFIED PLATFORM --------------------------------------
372 ##-------------------------------- DEFAULT APPLICATION --------------------------------------
373 APPLICATIONTMP="$(mktemp smpitmp-appXXXXXX)"
374 #APPLICATIONTMP="app.xml"
376 cat > ${APPLICATIONTMP} <<APPLICATIONHEAD
377 <?xml version='1.0'?>
378 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd">
379 <platform version="4.1">
382 ##---- cache hostnames of hostfile---------------
383 if [ -n "${HOSTFILE}" ] && [ -f ${HOSTFILE} ]; then
384 hostnames=$(cat ${HOSTFILE} | tr '\n\r' ' ')
387 if [ "${EXTOPT}" = "smpi_replay" ]; then
388 APP_TRACES=$PROC_ARGS
389 if [ -n "${APP_TRACES}" ] && [ -f "${APP_TRACES}" ]; then
390 hosttraces=$(cat ${APP_TRACES} | tr '\n\r' ' ' )
391 NUMTRACES=$(cat ${APP_TRACES} | wc -l)
394 printf "File not found: %s\n" "${APP_TRACES:-\${APP_TRACES\}}" >&2
399 ##----------------------------------------------------------
400 ## generate application.xml with hostnames from hostfile:
401 ## the name of host_i (1<=i<=p, where -np p) is the line i in hostfile (where -hostfile hostfile), or "host$i" if
402 ## hostfile has less than i lines.
403 ##----------------------------------------------------------
405 HAVE_SEQ="$(which seq 2>/dev/null)"
407 if [ -n "${HAVE_SEQ}" ]; then
408 SEQ=$(${HAVE_SEQ} 0 $(( NUMPROCS - 1)))
411 while [ $cnt -lt ${NUMPROCS} ] ; do
419 ##---- generate <actor> tags------------------------------
420 #prepare arguments at once
421 for ARG in $PROC_ARGS; do
422 XML_ARGS="${XML_ARGS}""<argument value=\"${ARG}\"/>
428 j=$(( i % hostfile_procs + 1 ))
429 host=$(eval "echo \${$j}")
431 ##---- optional display of ranks to actor mapping
432 if [ ${MAPOPT} = 1 ]; then
433 echo "[rank $i] -> $host"
436 echo " <actor host=\"${host}\" function=\"$i\"> <!-- function name used only for logging -->
437 <argument value=\"smpirun\"/> <!-- instance -->
438 <argument value=\"$i\"/> <!-- rank -->" >> ${APPLICATIONTMP}
439 if [ ${REPLAY} = 1 ]; then
440 if [ ${NUMTRACES} -gt 1 ]; then
441 echo " <argument value=\"$(echo $hosttraces|cut -d' ' -f$j)\"/>" >> ${APPLICATIONTMP}
443 echo " <argument value=\"$(echo $hosttraces|cut -d' ' -f1)\"/>" >> ${APPLICATIONTMP}
446 echo ${XML_ARGS} >> ${APPLICATIONTMP}
448 echo " </actor>" >> ${APPLICATIONTMP}
451 cat >> ${APPLICATIONTMP} <<APPLICATIONFOOT
454 ##-------------------------------- end DEFAULT APPLICATION --------------------------------------
455 ##---------------------- SMPI TRACING OPTIONS ---------------------------------
456 if [ -n "${TRACE_ACTIVE}" ]; then
457 #define trace filename
458 if [ -n "${TRACE_TI_ACTIVE}" ]; then
459 if [ -z "${TRACE_FILENAME}" ]; then
460 TRACE_FILENAME="smpi_simgrid.txt"
462 TRACEOPTIONS="--cfg=tracing:yes --cfg=tracing/filename:${TRACE_FILENAME} --cfg=tracing/smpi:yes --cfg=tracing/smpi/format:TI --cfg=tracing/smpi/computing:yes"
464 if [ -z "${TRACE_FILENAME}" ]; then
465 TRACE_FILENAME="smpi_simgrid.trace"
467 TRACEOPTIONS="--cfg=tracing:yes --cfg=tracing/filename:${TRACE_FILENAME} --cfg=tracing/smpi:yes"
470 if [ -n "${TRACE_COMMENT}" ]; then
471 TRACEOPTIONS="${TRACEOPTIONS} --cfg=tracing/comment:${TRACE_COMMENT}"
474 if [ -n "${TRACE_COMMENT_FILE}" ]; then
475 TRACEOPTIONS="${TRACEOPTIONS} --cfg=tracing/comment-file:${TRACE_COMMENT_FILE}"
478 if [ -n "${TRACE_GROUPED}" ]; then
479 TRACEOPTIONS="${TRACEOPTIONS} --cfg=tracing/smpi/group:yes"
482 if [ -n "${TRACE_RESOURCE}" ]; then
483 TRACEOPTIONS="${TRACEOPTIONS} --cfg=tracing/categorized:yes --cfg=tracing/uncategorized:yes"
486 if [ -n "${TRACE_VIVA}" ]; then
487 TRACEOPTIONS="${TRACEOPTIONS} --cfg=viva/categorized:smpi_cat.plist --cfg=viva/uncategorized:smpi_uncat.plist"
490 ##---------------------- end SMPI TRACING OPTIONS ---------------------------------
492 export SMPI_GLOBAL_SIZE=${NUMPROCS}
493 if [ -n "${KEEP}" ] ; then
494 echo ${EXEC} ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP}
495 if [ ${HOSTFILETMP} = 1 ] ; then
496 echo "Generated hostfile ${HOSTFILE} kept."
498 if [ ${UNROLLEDHOSTFILETMP} = 1 ] ; then
499 echo "Generated unrolled hostfile ${UNROLLEDHOSTFILE} kept."
503 # Execute the process
505 # The shell still need to be alive for the duration in order to do some cleanup after the process.
507 # We are going through great lengths in order to both keep stdin and be able to handle signals:
509 # * The job is launched in the background in order to be able to handle signals.
511 # * The FD 3 is used to temporarily store FD 1. This is because the shell connects FD 1 to /dev/null when the command
512 # is launched in the background: this can be overriden in bash but not in standard bourne shell.
514 ${WRAPPER} "@SMPIMAIN@" ${EXEC} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP} <&3 3>&- &
521 # Keep temporary files on failures to help debugging
523 if [ ${status} -ne 0 ] ; then
524 if [ -z "${KEEP}" ]; then
525 echo ${EXEC} ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP}
526 if [ ${HOSTFILETMP} = 1 ] ; then
527 echo "Generated hostfile ${HOSTFILE} kept."
529 if [ ${UNROLLEDHOSTFILETMP} = 1 ] ; then
530 echo "Generated unrolled hostfile ${UNROLLEDHOSTFILE} kept."
533 echo "Execution failed with code ${status}."