Using ACS Cluster: Diferență între versiuni

Versiunea de la data 26 aprilie 2018 16:02

Steps:

copy files on the cluster (scp -r <path to local folder> <moodle id>@fep.grid.pub.ro:~)
log in with moodle id (ssh <moodle id>@fep.grid.pub.ro)
log in gpu queue (qlogin -q ibm-dp.q)
module load 1 (module load libraries/cuda)
module load 2 (module load utilities/opencl)
compile opencl project (make)
get back to fep (exit)
submit job on gpgpu queue for execution (qsub -q ibm-dp.q -b y <path to executable>)
wait for the job to finish (qstat -j <job id>)
check the files made in local directory (<executable name>.o/e<job id>)

For more info see:

[ACS ASC Lab]: cs.curs.pub.ro/wiki/asc/asc:lab7:index

@@ Linia 1: / Linia 1: @@
 Steps:
-: copy files on the cluster (scp -r <path to local folder> <moodle id>@fep.grid.pub.ro:~)
+# copy files on the cluster (scp -r <path to local folder> <moodle id>@fep.grid.pub.ro:~)
-: log in with moodle id (ssh <moodle id>@fep.grid.pub.ro)
+# log in with moodle id (ssh <moodle id>@fep.grid.pub.ro)
-: log in gpu queue (qlogin -q ibm-dp.q)
+# log in gpu queue (qlogin -q ibm-dp.q)
-: module load 1 (module load libraries/cuda)
+# module load 1 (module load libraries/cuda)
-: module load 2 (module load utilities/opencl)
+# module load 2 (module load utilities/opencl)
-: compile opencl project (make)
+# compile opencl project (make)
-: get back to fep (exit)
+# get back to fep (exit)
-: submit job on gpgpu queue for execution (qsub -q ibm-dp.q -b y <path to executable>)
+# submit job on gpgpu queue for execution (qsub -q ibm-dp.q -b y <path to executable>)
-: wait for the job to finish (qstat -j <job id>)
+# wait for the job to finish (qstat -j <job id>)
-: check the files made in local directory (<executable name>.o/e<job id>)
+# check the files made in local directory (<executable name>.o/e<job id>)
 For more info see:
 [ACS ASC Lab]: cs.curs.pub.ro/wiki/asc/asc:lab7:index
 [Cluster cheat sheet]: http://cs.curs.pub.ro/wiki/asc/_media/asc:resurse:cluster-cheat-sheet.pdf