Intel® DevCloud
Help for those needing help starting or connecting to the Intel® DevCloud
1624 Discussions

DevCloud DPCPP Installation

RAror6
New Contributor I
1,560 Views

Hello,

I am running DPC++ on DevCloud but for some reason it says cannot find DPCPP when I run my application. How can I fix this issue?

I believe I have run a script on DevCloud that might have messed up my installation of oneapi and dpcpp. Commands like 'oneapi-cli' don't work as well until I run the setvars.sh file.

Thank you!

0 Kudos
6 Replies
AbhishekD_Intel
Moderator
1,560 Views

Hi,

Can you please share with us what is the exact error that you are getting while running your application with DPCPP?

Please share with us details of the script that you think could have messed up your OneAPI & DPCPP, and also please share the error logs that you are getting while using oneapi-cli.

 

You may also try compiling your application and oneapi-cli command on the compute node. For using oneapi-cli you have to be on the compute node, you may try interactive mode (-I) while using qsub to select a compute node.

 

Warm Regards,

Abhishek

 

 

0 Kudos
RAror6
New Contributor I
1,560 Views

Thank you for the reply!

I am getting the error or "dpcpp: not found". I believe the script that messed this up was in the GettingStarted code for oneDNN. There is a script to remove "setvars.sh" form bash profile for custom oneDNN tensorflow.

In my command line I now necessarily need to type "source setvars.sh" to run dpcpp. But in Jupyter notebooks my kernel is not recognizing dpcpp as a valid command.

I have tried my best to get it running so far, but cannot.

0 Kudos
AbhishekD_Intel
Moderator
1,560 Views

Hi,

Try selecting compute node in interactive mode after logged in into DevCloud you can follow the below command:

qsub -I -l nodes=1:ppn=2

And then try executing your commands like oneapi-cli and try running your application with DPCPP on selected compute node. If the error still persists then let us know, it should resolve your issue.

And please give us more details regarding the script which is asking you to remove "setvarsh.sh" or give us details of steps that you followed and encounter this error. Because we tried building oneDNN_Getting_Started example but we didn't get this error. So please give us details of the steps that you followed so that we can reproduce this issue.

 

Warm Regards,

Abhishek

0 Kudos
RAror6
New Contributor I
1,560 Views

Hello,

Thank you for the reply!

I tried running it in the interactive mode but it seems to time out.

Please find attached a text file containing the code that started the errors. This is the same code that is found in the Jupyter Notebook for "getting_Started" in oneDNN.

In more detail:
1) My python kernel in Junpyter Notebook could recognise "! dpcpp" as a valid command
2) I believe that after running the attached script, my Jupyter Notebook kernel doesn't recognise the bash commands and dpcpp
3) In my terminal on DevCloud, my oneAPI kernel is not default now too. I need to run the setvars.sh script in order to run oneAPI and any of it's libraries.

It is always possible that something else caused this error, but in my opinion it is this script that did.


Thank you!

0 Kudos
AbhishekD_Intel
Moderator
1,552 Views

Hi,

Pardon us for the delay and inconvenience. We have looked into the problem and we will suggest you replace you bash_profile with the one attached with this post and source it or re-login to devcloud and check whether you are having the same problem.

Please update us with the errors and problems if any.

Find the bash_profile here:

# .bash_profile

# Get the aliases and functions
if [ -f ~/.bashrc ]; then
   . ~/.bashrc
fi

# User specific environment and startup programs
export PATH=$PATH:$HOME/.local/bin:$HOME/bin

# Enable Intel tools
export INTEL_LICENSE_FILE=/usr/local/licenseserver/psxe.lic
export PATH=/glob/intel-python/python3/bin/:/glob/intel-python/python2/bin/:${PATH}
source /glob/development-tools/parallel-studio/bin/compilervars.sh intel64
export PATH=$PATH:/bin
if [ -d /opt/intel/inteloneapi ]; then source /opt/intel/inteloneapi/setvars.sh > /dev/null 2>&1; fi

# Make sure that most programs (in particular, pip) leave temp files locally
if [ ! -d ${HOME}/tmp ]; then
  mkdir ${HOME}/tmp
fi
export TMPDIR=${HOME}/tmp
#export PBS_DEFAULT=v-qsvr-nda

 

 

Warm Regards,

Abhishek

 

0 Kudos
AbhishekD_Intel
Moderator
1,490 Views

Hi,

Please give us an update on this thread, so that we can dig into your problem.


Thank you.


0 Kudos
Reply