Running on HPC cluster

[CirclestheSecond]

Hello,

I have been using DSPH for about 4+ years on my personal computer, recently I have been trying to run it on HPC nodes and I keep getting an error when it tries to execute dualsphysicsgpu, GenCase works without a problem and I can see the initial state of my model.

After Gencase when it tries to load DsphConfig, it shows an error message which reads

LoadDsphConfig> .../bin/linux/DsphConfig.xml

*** Exception (JSphCfgRun::ErrorParm) at JCfgRunBase.cpp:137
Text: Parameter "name_out" unrecognised or invalid. (Level cfg:0, Parameter:3)

Based on a few suggestions on the forum, I have changed the makefile to match with the cuda version and compute capability of HPC gpu. I built the executables successfully, but I am not able to execute dualsphysicsgpu.

Since the error appears after DsphConfig, I thought maybe it had a problem accessing the name_out folder, I then modified my ~/.bashrc to point to the right directory. But still no success.

I was hoping someone had a similar problem and know how to fix it.

[CirclestheSecond]

Use this following line to execute dualsphysicsgpu and it works

${dualsphysicsgpu} -gpu :0 ${dirout}/${name} ${dirout} -dirdataout data -svres

${dualsphysicsgpu} -gpu :0 -name ${dirout}/${name} -dirout ${dirout} -dirdataout data -svres
if [ $? -ne 0 ] ; then fail; fi