- Infos im HLRS Wiki sind nicht rechtsverbindlich und ohne Gewähr -
- Information contained in the HLRS Wiki is not legally binding and HLRS is not responsible for any damages that might result from its use -
Vampirtrace: Difference between revisions
Line 85: | Line 85: | ||
== Trouble shooting == | == Trouble shooting == | ||
If the program crashes unexpectedly it is possible that there is not enough space on the node local /tmp. A solution is to use a workspace in combination with the VT_PROFM_LDIR variable | If the program crashes unexpectedly it is possible that there is not enough space on the node local /tmp. A solution is to use a workspace in combination with the <tt>VT_PROFM_LDIR</tt> variable | ||
In some cases problems concerning library issues occure. In this case it could help | In some cases problems concerning library issues occure. In this case it could help |
Revision as of 17:36, 26 February 2010
The VampirTrace library allows MPI communication events of a parallel program to be recorded as a trace file. Additionally, certain program-specific events can also be included. To record MPI communication events, simply relink the program with the VampirTrace library. A new compilation of the program source code is only necessary if program-specific events should be added.
Introduction
Tracing with Vampir in three steps:
- Compiling with the vampirtrace wrappercompiler
- Running the application
- Analysing the trace files using vampir or vampir-server
For detailed usage of Vampirtrace see the user manual which can either be found on your local machine or in the internet.
Usage
Compiling with the vampirtrace wrappercompiler
First load the needed software module:
Now you can compile your application using the wrapper compilers vtcc, vtc++ or vtf90 as described in the user VampirTrace manual.
vtf90 -vt:f90 mpif90 vtcc -vt:cc mpicc
Generating the trace files
Run your applicatin as usual. This will generate the needed tracefiles
BW-Grid special:
On the BW-Grid you first have to login to a node as some libs are not accessible from the frontend
Viewing traces with vampir or Vampir-Server
There are two GUIs available to view your trace files: Vampir and Vampir-Server. For both you will need read access to all generated output files (*.def.z, *.events.z, *.otf).
Vampir
Vampir is the sequential version of the graphical trace file analyzing software. It is capable of analyzing smaller trace files and should not be used for files which have 100MB or more.
Vampir-Server
Vampir-Server is the parallel version of the graphical trace file analyzing software. It can be found on the SX-8 frontend (a1.hww.de). The tool consists of a client and a server. The server itself is a parallel MPI Program.
First load the necessary software module
Now start the server and afterwards the client
Remember the port number which will be displays as you will need it later on. Then start the graphical client
The last step is to connect the client to the server using the port number offered you from the server. This is done under
file <math>\rightarrow</math> Connect to Server ...
Then you can open your traces via
file <math>\rightarrow</math> Open Tracefile ...
Select the desired *.otf file
Trouble shooting
If the program crashes unexpectedly it is possible that there is not enough space on the node local /tmp. A solution is to use a workspace in combination with the VT_PROFM_LDIR variable
In some cases problems concerning library issues occure. In this case it could help