Vampir on Cplant

About Vampir!

Vampir has two parts, the VampirTrace library which you link with your code on a compile node, and the Vampir client with which you analyse the results on your local desktop machine. You need to install the client on your local machine. At present, it does not run on the Cplant service or compile nodes.

Setup Vampir?

• Setup for VampirTrace on the Linux compile nodes:
  • The VampirTrace library is in /usr/local/share/vt. Add lines like the following to your Makefiles in order to link with the VampirTrace library:
    VAMPIR = -L/usr/local/share/vt/lib -lVT
MPI_HOME = $(CPLANT_PATH)
LIBS = $(VAMPIR) -L$(MPI_HOME)/lib -lpmpi -lmpi -lm

  • Set the following environment variable so that when you execute your code, VampirTrace will find the license:
    export PAL_LICENSEFILE=/cplant/vt/etc/license.dat

• Setup Vampir client on your local computer. To install in your home directory (e.g. $HOME):
  
  • Download the Vampir client to your local Linux workstation. The tar file is /usr/local/share/vt/Client.tar.gz.
  • Set the environment variable for the license file
    export PAL_LICENSEFILE=$HOME/Vampir/etc/license.dat
  • Execute the following lines to install in $HOME/Vampir.
    tar xzf Client.tar.gz
cd $HOME/Vampir
./install-Vampir

• You should be all set to go. Once you've run your code on Cplant, there will be a lot of trace files named codename.stf*. To analyse your code, move these files over to your local machine, and run vampir on them:
  vampir codename.stf

Where can I find additional documentation?

There's a simple example in /usr/local/share/vt/pingpong.

For VampirTrace, the best documentation is the man pages in /usr/local/share/vt/man. The client has it's own pdf documentation.
Tutorials, etc. can be found on the Pallas,GmBH web site