Differences

This shows you the differences between two versions of the page.

--- doku:mathematica [2015/05/13 13:32] – [VSC-3] ir
+++ doku:mathematica [2022/03/10 12:24] – [remote kernel connection] goldenberg
@@ Line 1: / Line 1: @@
 ======= Mathematica Batch Jobs =======
-===== VSC-2 =====
-==== Sun Grid Engine ====
-The simplest way is to create a file with your mathematica commands, e.g. 'math-input.m', and just input that in the batch file.
-<code>
-# the following option makes sure the job will run in the current directory
-#$ -cwd
-# this option makes sure the job has the same environment variables as the submission shell
-#$ -V
-MATHDIR=/opt/sw/Mathematica/90/bin
-$MATHDIR/math < math-input.m
-</code>
-If we name the above file as math-job.sh and place it in the same directory as math-input.m we can submit it from the same directory with
-<code>
-user@l01 $ qsub math-job.sh
-</code>
-The output will be left in file ''math-job.sh.o$JOB_ID'' and any errors in ''math-job.e$JOB_ID''.
-You could also incorporate mathematica commands in the job file itself, rather then have them in a separate file:
-<code>
-#$ -cwd
-#$ -V
-MATHDIR=/opt/sw/Mathematica/90/bin
-$MATHDIR/math <<END_MATH_COMMANDS
-+1
-*3
-END_MATH_COMMANDS
-</code>
-The above notations means that everything between "<<END_MATH_COMMANDS" and "END_MATH_COMMANDS" will be used as math program's input. You can again submit this job with qsub.
-==== Parallel Usage ====
-Jobscript:
-<code>
-#$ -N parallel
-#$ -q all.q
-#$ -pe mpich 32
-# $TMPDIR/machines shows which host machines are reserved for the current job;
-# copy this file to a temporary file which will be read in by Mathematica;
-# in the following the small Mathematics m-file will start a kernel on each machine;
-# since the kernel on the master is already running, it is sufficient
-# to start the kernels beginning with the second host in the machines file.
-cp -v $TMPDIR/machines machines_tmp
-/opt/sw/Mathematica/90/bin/math -run "<<math.m"
-</code>
-Mathematica script named ''math.m'':
-<code>
-(* configuration for starting remote kernels *)
-Needs["SubKernels`RemoteKernels`"]
-$RemoteCommand= "ssh `1` -n -l `3` \"/opt/sw/Mathematica/90/bin/math -mathlink -linkmode Connect `4` -linkname `2` -subkernel -noinit >& /dev/null &\""
-(* initialize the kernels on all machines defined in the host file *)
-hosts=Import["machines_tmp","List"]
-(* since the kernel on the master is already running, initialization starts with host 2 *)
-imin=2;
-imax=Length[hosts];
-idelta=1;
-Do[
-	Print["starting Kernel: ",i," on ",hosts[[i]]];
-	LaunchKernels[RemoteMachine[hosts[[i]]]];,
-	{i,imin,imax,idelta}
-]
-(* actual calculation *)
-primelist = ParallelTable[Prime[k], {k, 1, 20000000}];
-Print[primelist]
-</code>
-===== VSC-3 =====
 === SLURM ===
 In order to be able to use Mathematica, you have to load the program with the ''module load xyz'' command
@@ Line 109: / Line 22: @@
 Mathematica script example ''math-vsc3.m'':
 <code>
-(* configuration for starting remote kernels *)
+GetEnvironment["MATH_BIN"]
+math = Environment["MATH_BIN"]
+kernelsperhost = 16
+hosts=Import["nodelist","List"];
 Needs["SubKernels`RemoteKernels`"]
+$RemoteCommand = "ssh -x -f -l `3` `1` " <> math <> " -mathlink -linkmode Connect `4` -linkname '`2`' -subkernel -noinit"
-(* initialize the kernels on all machines defined in the host file *)
+(* first host with one kernel less *)
+LaunchKernels[RemoteMachine[hosts[[1]], kernelsperhost-1]];
-hosts=Import["machines_tmp","List"]
-(* on the master node initialize only one kernel less, since one is already running *)
 imin=2;
 imax=Length[hosts];
 idelta=1;
 Do[
-        Print["starting Kernel: ",i," on ",hosts[[i]]];
+        LaunchKernels[RemoteMachine[hosts[[i]], kernelsperhost]];
-        LaunchKernels[RemoteMachine[hosts[[i]]]];,
+        , {i, imin, imax, idelta}]
-        {i,imin,imax,idelta}
-]
+primelist = ParallelTable[Prime[k], {k, 1, 20000000}];
-(* actual calculation *)
-primelist = ParallelTable[Prime[k], {k, 1, 2000}];
 Print[primelist]
 </code>
 sbatch job script
 ''jobPar.sh'':
@@ Line 139: / Line 54: @@
 #!/bin/bash
 #
-#SBATCH -J par                      # job name
+#SBATCH -J par
-#SBATCH -N 2                        # number of nodes=2
+#SBATCH -N 2
-#SBATCH --ntasks-per-node=16        # uses all cpus of one node
+#SBATCH -L mathematica@vsc
-#SBATCH --ntasks-per-core=1
-#SBATCH --threads-per-core=1
-module load Mathematica/10.0.2 # load desired version
-scontrol show hostnames $SLURM_NODELIST > ./nodelist
+module purge
+module load Mathematica/10.0.2 # load desired version
-rm machines_tmp
-tasks_per_node=16
+export MATH_BIN=`which math`
-nodes=2
+#export MATH_PROC=16
-for ((line=1; line<=nodes; line++))
+scontrol show hostnames $SLURM_NODELIST > nodelist
-do
+#execute prolog for getting access to license everywhere
-    for ((tN=1; tN<=tasks_per_node; tN++))
+srun hostname
-    do
-        head -$line nodelist | tail -1 >> machines_tmp
-    done
-done
-echo "this is the content of the host machines file"
-cat machines_tmp
 math -run < math-vsc3.m
@@ Line 171: / Line 77: @@
 [username@l32]$ squeue -u username    # check state of your job
 </code>
+==== remote kernel connection ====
+The setup follows
+[[https://cc-mathematik.univie.ac.at/services/vsc3-cluster/remote-mathematica-kernel/|Local mathematica with remote kernel]]
+with the
+**Shell command to launch kernel:**
+for VSC3+
+<code>
+/Users/<LOCAL-USERNAME>/Library/tunnel.sh <VSC-USERNAME>@localhost:9998 /opt/sw/x86_64/generic/Mathematica/11.3/bin/math `linkname`
+</code>
+for VSC4
+<code>
+/Users/<LOCAL-USERNAME>/Library/tunnel.sh <VSC-USERNAME>@localhost:9998 /opt/sw/vsc4/VSC/x86_64/generic/Mathematica/12.3.1/bin/math `linkname`
+</code>
+Note that on VSC4, the salloc command [as described on the
+cc-mathematik.univie.ac.at web link] does not need the Mathematica
+license. Just leave the -L parameter away.