X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/c42c8f1c9c48a271f8f5e8af6e7e9809ddfa1864..fbcf6ab31cae1988be858f9f894dafe529c575d7:/src/smpi/smpirun.in

diff --git a/src/smpi/smpirun.in b/src/smpi/smpirun.in
index 8e07eaf1f9..5fbc9b3511 100755
--- a/src/smpi/smpirun.in
+++ b/src/smpi/smpirun.in
@@ -1,7 +1,6 @@
-#! /bin/sh
+#!/usr/bin/env sh
 
-# Copyright (c) 2007-2015. The SimGrid Team.
-# All rights reserved.
+# Copyright (c) 2007-2019. 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.
@@ -24,27 +23,33 @@ NETWORK_BANDWIDTH="${DEFAULT_NETWORK_BANDWIDTH}"
 NETWORK_LATENCY="${DEFAULT_NETWORK_LATENCY}"
 SPEED="${DEFAULT_SPEED}"
 
-SIMOPTS="--cfg=surf/precision:1e-9 --cfg=network/model:SMPI --cfg=network/TCP-gamma:4194304"
+PRIVATIZE="--cfg=smpi/privatization:${SMPI_PRIVATIZATION:-@HAVE_PRIVATIZATION@}"
+
+SIMOPTS="--cfg=surf/precision:1e-9 --cfg=network/model:SMPI"
 
 #usage to print the way this script should be called
 usage () {
     cat <<EOF
 Usage: $0 [OPTIONS] -platform <xmldesc> -hostfile <hostfile> program [program-options]
+       $0 [OPTIONS] -platform <xmldesc> -hostfile <hostfile> -replay <tracefile> [program] [program-options]
 Options:
   -keep-temps                # don't remove the generated files after execution
-  -wrapper <command>         # use command to run the program (e.g. "valgrind")
+  -wrapper <command>         # 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 <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)
   -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
   -trace-grouped             # group MPI processes by location
   -trace-resource            # trace resource utilization
-  -trace-viva                # generate configuration for Viva's GraphView
   -trace-file <tracefile>    # name of the tracefile (simgrid_smpi.trace)
-  -ext <value>               # additional parameter (reserved)
+  -replay <tracefile>        # replays a trace instead of actually executing an application
 
   -version                   # Displays the SimGrid version (human readable)
   -git-version               # Displays the git hash of SimGrid
@@ -62,15 +67,21 @@ then
     exit
 fi
 
-EXTOPT=""
 WRAPPER=""
 HOSTFILE=""
 HOSTFILETMP=0
+MAPOPT=0
+REPLAY=0
 
 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
+    exit 1
+}
+
 smpirun_cleanup()
 {
   if [ -z "${KEEP}" ] ; then
@@ -134,33 +145,27 @@ while true; do
         "-platform")
 	    PLATFORM="$2"
             if [ ! -f "${PLATFORM}" ]; then
-		echo "[`basename $0`] ** error: the file '${PLATFORM}' does not exist. Aborting."
-		exit 1
-            fi
-	    shift 2
-            ;;
-        "-hostfile")
-	    HOSTFILE="$2"
-            if [ ! -f "${HOSTFILE}" ]; then
-		echo "[`basename $0`] ** error: the file '${HOSTFILE}' does not exist. Aborting."
-		exit 1
+                die "the file '${PLATFORM}' does not exist"
             fi
 	    shift 2
             ;;
-        "-machinefile")
+        "-hostfile" | "-machinefile")
 	    HOSTFILE="$2"
             if [ ! -f "${HOSTFILE}" ]; then
-		echo "[`basename $0`] ** error: the file '${HOSTFILE}' does not exist. Aborting."
-		exit 1
+                die "the file '${HOSTFILE}' does not exist"
             fi
 	    shift 2
             ;;
-        "-ext")
-            EXTOPT="$2"
+        "-replay")
+            APP_TRACES="$2"
             shift 2
             ;;
+	"-no-privatize")
+	    PRIVATIZE="--cfg=smpi/privatization:no"
+	    shift 1
+	    ;;
         "-map")
-	    MAPOPT="true"
+	    MAPOPT=1
             shift 1
             ;;
         "-trace")
@@ -192,18 +197,33 @@ while true; do
             TRACE_RESOURCE="true"
             shift 1
             ;;
-        "-trace-viva")
-   	    TRACE_VIVA="true"
-   	    shift 1
-            ;;
         "-keep-temps")
-	    KEEP="true"
+            KEEP="true"
+            SIMOPTS="$SIMOPTS --cfg=smpi/keep-temps:yes"
             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
@@ -233,12 +253,25 @@ while true; do
     esac
 done
 
-if [ -n "$WRAPPER" ]; then
-    EXEC="$WRAPPER $1"
+if [ -n "${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
 fi
-shift
 
 # steel --cfg and --logs options
 while [ $# -gt 0 ]; do
@@ -252,7 +285,7 @@ while [ $# -gt 0 ]; do
             ;;
         *)
             PROC_ARGS="${PROC_ARGS:+$PROC_ARGS }$1"
-            shift      
+            shift
             ;;
     esac
 done
@@ -267,25 +300,27 @@ if [ -z "${HOSTFILE}" ] ; then
     HOSTFILETMP=1
     HOSTFILE="$(mktemp smpitmp-hostfXXXXXX)"
     perl -ne 'print "$1\n" if /.*<host.*?id="(.*?)".*?\/>.*/' ${PLATFORM} > ${HOSTFILE}
-    perl -ne 'if (/.*<cluster.*?prefix="(.*?)".*?radical="(.*?)".*?suffix="(.*?)".*/) { 
-                my ($pre,$rad,$post)=($1,$2,$3); 
-	        for my $elm (split(",",$rad)) { 
-	          if ($elm=~/^([^-]*?)-([^-]*)$/) { 
-	             for (my $i=$1; $i<=$2;$i++) { 
-	                print "$pre$i$post\n"; 
+    # put all <cluster tag on its own line.
+    cat ${PLATFORM} | tr '\n' ' ' | sed 's/<cluster/\n<cluster/' | \
+      perl -ne 'if (m/.*<cluster.*?prefix="(.*?)".*?radical="(.*?)".*?suffix="(.*?)".*/s) {
+                my ($pre,$rad,$post)=($1,$2,$3);
+	        for my $elm (split(",",$rad)) {
+	          if ($elm=~/^([^-]*?)-([^-]*)$/) {
+	             for (my $i=$1; $i<=$2;$i++) {
+	                print "$pre$i$post\n";
 	             }
 	          } else {
 	             print "$pre$elm$post\n";
 	          }
 	        }
 	      } elsif (/<cluster/) {
-	        die ("Unparsable cluster tag. Either provide an hostfile yourself or give the attributes prefix, radical and suffix in that order on the <cluster line");
+            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.");
 	      }' ${PLATFORM} >> ${HOSTFILE}
 fi
 UNROLLEDHOSTFILETMP=0
 
 #parse if our lines are terminated by :num_process
-multiple_processes=`grep -c ":" $HOSTFILE`
+multiple_processes=$(grep -c ":" $HOSTFILE)
 if [ "${multiple_processes}" -gt 0 ] ; then
     UNROLLEDHOSTFILETMP=1
     UNROLLEDHOSTFILE="$(mktemp smpitmp-hostfXXXXXX)"
@@ -298,10 +333,9 @@ if [ "${multiple_processes}" -gt 0 ] ; then
 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]" $HOSTFILE`
+hostfile_procs=$(grep -c "[a-zA-Z0-9]" $HOSTFILE)
 if [ ${hostfile_procs} = 0 ] ; then
-   echo "[`basename $0`] ** error: the hostfile '${HOSTFILE}' is empty. Aborting." >&2
-   exit 1
+    die "the hostfile '${HOSTFILE}' is empty"
 fi
 
 if [ -z "${NUMPROCS}" ] ; then
@@ -310,7 +344,7 @@ if [ -z "${NUMPROCS}" ] ; then
 fi
 
 if [ ${NUMPROCS} -gt ${hostfile_procs} ] ; then
-    echo "You requested to use ${NUMPROCS} processes, but there is only ${hostfile_procs} processes in your hostfile..." >&2
+    echo "You requested to use ${NUMPROCS} ranks, but there is only ${hostfile_procs} processes in your hostfile..." >&2
 fi
 
 ##-------------------------------- DEFAULT or SPECIFIED PLATFORM --------------------------------------
@@ -319,9 +353,9 @@ if [ -z "${PLATFORM}" ]; then
 
     cat > ${PLATFORMTMP} <<PLATFORMHEAD
 <?xml version='1.0'?>
-<!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd">
-<platform version="4">
-<AS id="AS0" routing="Full">
+<!DOCTYPE platform SYSTEM "https://simgrid.org/simgrid.dtd">
+<platform version="4.1">
+<zone id="AS0" routing="Full">
 PLATFORMHEAD
 
     i=${NUMPROCS}
@@ -347,7 +381,7 @@ PLATFORMHEAD
     done
 
     cat >> ${PLATFORMTMP} <<PLATFORMFOOT
-</AS>
+</zone>
 </platform>
 PLATFORMFOOT
 
@@ -361,21 +395,20 @@ APPLICATIONTMP="$(mktemp smpitmp-appXXXXXX)"
 
 cat > ${APPLICATIONTMP} <<APPLICATIONHEAD
 <?xml version='1.0'?>
-<!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd">
-<platform version="4">
+<!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 ${HOSTFILE} | tr '\n\r' '  ')
-    NUMHOSTS=$(cat ${HOSTFILE} | wc -l)
 fi
 
-if [ "${EXTOPT}" = "smpi_replay" ]; then
-    APP_TRACES=$PROC_ARGS
-    if [ -n "${APP_TRACES}" ] && [ -f "${APP_TRACES}" ]; then
+if [ -n "${APP_TRACES}" ]; then
+    if [ -f "${APP_TRACES}" ]; then
         hosttraces=$(cat ${APP_TRACES} | tr '\n\r' '  ' )
         NUMTRACES=$(cat ${APP_TRACES} | wc -l)
+        REPLAY=1
     else
         printf "File not found: %s\n" "${APP_TRACES:-\${APP_TRACES\}}" >&2
         exit 1
@@ -384,14 +417,14 @@ fi
 
 ##----------------------------------------------------------
 ##  generate application.xml with hostnames from hostfile:
-##  the name of host_i (1<=i<=p, where -np p) is the line i in hostfile (where -hostfile hostfile), or "host$i" if 
+##  the name of host_i (1<=i<=p, where -np p) is the line i in hostfile (where -hostfile hostfile), or "host$i" if
 ##  hostfile has less than i lines.
 ##----------------------------------------------------------
 
-HAVE_SEQ="`which seq 2>/dev/null`"
+HAVE_SEQ="$(which seq 2>/dev/null)"
 
 if [ -n "${HAVE_SEQ}" ]; then
-    SEQ=`${HAVE_SEQ} 0 $((${NUMPROCS}-1))`
+    SEQ=$(${HAVE_SEQ} 0 $(( NUMPROCS - 1)))
 else
     cnt=0
     while [ $cnt -lt ${NUMPROCS} ] ; do
@@ -400,37 +433,39 @@ else
     done
 fi
 
-##---- generate <process> tags------------------------------
+set -- $hostnames
+
+##---- generate <actor> tags------------------------------
+#prepare arguments at once
+for ARG in $PROC_ARGS; do
+  XML_ARGS="${XML_ARGS}""<argument value=\"${ARG}\"/>
+"
+done
+
 for i in ${SEQ}
 do
-    if [ -n "${HOSTFILE}" ]; then
-	j=$(( $i % ${NUMHOSTS} + 1 ))
-    fi
-    ##---- optional display of ranks to process mapping
-    if [ -n "${MAPOPT}" ]; then
-	echo "[rank $i] -> $(echo $hostnames|cut -d' ' -f$j)"
-    fi
+    j=$(( i % hostfile_procs + 1 ))
+    host=$(eval "echo \${$j}")
 
-    if [ -z "$(echo $hostnames|cut -d' ' -f$j)" ]; then
-	host="host"$($j)
-    else
-	host="$(echo $hostnames|cut -d' ' -f$j)"
+    ##---- optional display of ranks to actor mapping
+    if [ ${MAPOPT} = 1 ]; then
+      echo "[rank $i] -> $host"
     fi
-    echo "  <process host=\"${host}\" function=\"$i\"> <!-- function name used only for logging -->" >> ${APPLICATIONTMP}
-    echo "    <argument value=\"1\"/> <!-- instance -->" >> ${APPLICATIONTMP}
-    echo "    <argument value=\"$i\"/> <!-- rank -->" >> ${APPLICATIONTMP}
-    if [ "${EXTOPT}" = "smpi_replay" ]; then
+
+    echo "  <actor host=\"${host}\" function=\"$i\"> <!-- function name used only for logging -->
+    <prop id=\"instance_id\" value=\"smpirun\"/>
+    <prop id=\"rank\" value=\"$i\"/>" >> ${APPLICATIONTMP}
+    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)\"/>" >> ${APPLICATIONTMP}
         else
             echo "    <argument value=\"$(echo $hosttraces|cut -d' ' -f1)\"/>" >> ${APPLICATIONTMP}
         fi
-    else 
-        for ARG in $PROC_ARGS; do
-            echo "    <argument value=\"${ARG}\"/>" >> ${APPLICATIONTMP}
-        done
+    else
+    echo ${XML_ARGS} >> ${APPLICATIONTMP}
     fi
-    echo "  </process>" >> ${APPLICATIONTMP}
+    echo "  </actor>" >> ${APPLICATIONTMP}
 done
 
 cat >> ${APPLICATIONTMP} <<APPLICATIONFOOT
@@ -467,21 +502,18 @@ if [ -n "${TRACE_ACTIVE}" ]; then
     if [ -n "${TRACE_RESOURCE}" ]; then
         TRACEOPTIONS="${TRACEOPTIONS} --cfg=tracing/categorized:yes --cfg=tracing/uncategorized:yes"
     fi
-
-    if [ -n "${TRACE_VIVA}" ]; then
-        TRACEOPTIONS="${TRACEOPTIONS} --cfg=viva/categorized:smpi_cat.plist --cfg=viva/uncategorized:smpi_uncat.plist"
-    fi
 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} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP}
+    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." 
+        echo "Generated unrolled hostfile ${UNROLLEDHOSTFILE} kept."
     fi
 fi
 
@@ -496,13 +528,35 @@ fi
 # * 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.
 exec 3<&0
-${EXEC} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP} <&3 3>&- &
+${WRAPPER} "@SMPIMAIN@" ${EXEC} ${PRIVATIZE} ${TRACEOPTIONS} ${SIMOPTS} ${PLATFORMTMP} ${APPLICATIONTMP} <&3 3>&- &
 pid=$!
 exec 3>&-
 wait $pid
 status=$?
+# With dash on Windows WSL/Ubuntu, "wait" sometimes returns early with an exit
+# status of 128. Try again.
+while test $status -eq 128 && kill -0 $pid 2>/dev/null; do
+    wait $pid
+    status=$?
+done
 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 [ ${HOSTFILETMP} = 1 ] ; then
+            echo "Generated hostfile ${HOSTFILE} kept."
+        fi
+        if [ ${UNROLLEDHOSTFILETMP} = 1 ] ; then
+            echo "Generated unrolled hostfile ${UNROLLEDHOSTFILE} kept."
+        fi
+    fi
+    echo "Execution failed with code ${status}."
+    KEEP=true
+fi
+
 smpirun_cleanup
 
 exit $status