vampir OverviewVampir is a software performance visualizer focused on highly parallel applications. It presents a unified view on an application run including the use of programming paradigms like MPI, OpenMP, PThreads, CUDA, OpenCL and OpenACC. It also incorporates file I/O, hardware performance counters and other performance data sources. Various interactive displays offer detailed insight into the performance behavior of the analyzed application. Vampir's scalable analysis server and visualization engine enable interactive navigation of large amounts of performance data. Score-P and TAU generate OTF2 trace files for Vampir to visualize.
UsageThere are three ways to use Vampir. First, you can install and use Vampir on your own computer. Second, you can use Vampir on OLCF systems via X-forwarding. Third, if you want to use Vampir's powerful HPC-enabled analysis server, you can launch Vampir as a job and connect to it from your local computer.
Install Vampir locally
Vampir's installers are in directory
/sw/sources/vampir/client on any of the OLCF systems. Download the installer for your operating system from there. Follow the instructions of the installer. When running Vampir for the first time, you are asked to activate your copy. For this use the license
/sw/sources/vampir/vampir.license and make sure to use your ORNL email address for the process.
Without an ORNL email address, you can use the license
/sw/sources/vampir/vampir-remote.license. With this license the Vampir client can only connect to Vampir server instances, for example launched on OLCF systems.
Use Vampir locallyIf you have a trace, copy it from the HPC system to your computer and open the .otf2-file via File -> Open. OTF2 traces have the following directory structure:
analysis-experiment-directory/ traces/ *.def *.evt traces.def traces.otf2
Use Vampir via X-ForwardingIf you do not install Vampir locally, you can log in to any OLCF system and use it from there. To enable X-forwarding, log in to the system using
ssh -X. Start Vampir via
module load vampir vampir &From here, use Vampir in the same way as if installed locally.
Use Vampir Client + ServerIf your local computer is too slow, runs out of memory, or you want to analyze large traces using an HPC system, Vampir's client+server approach is the right way. Loading the module
vampiron any OLCF system prints instructions on how to use Vampir in client+server mode.
$ module load vampir Run VampirServer via $ vampirserver start -- -AExecute the above port-forwarding command
[-q ] [-w ] and follow the instructions provided by this script. Run the Vampir GUI remotely via X-forwarding $ vampir & $ vampirserver start -- -A your_project_id Launching VampirServer... [..] User: your_user_name Password: random_password VampirServer <4192> listens on: nid06289:30081 Please run: ssh -L 30006:nid06289:30081 email@example.com on your desktop to create an ssh tunnel to VampirServer. Start Vampir on your desktop and choose 'Open Other -> Remote File' Description: titan, Server: localhost, Port: 30006 Authentication: Encrypted password Connection type: Socket Ignore 'More Options' User: your_user_name, Password: random_password
ssh -L ...on your local computer, start Vampir locally, and then connect to the launched server via Open Other -> Remote File.
$ module load vampir Run VampirServer via $ vampirserver start -- -PAt the moment, forwarding ports to Summit is not available. Therefore you have to start the Vampir client on the login node via
[-q ] [-w ] and follow the instructions provided by this script. Run the Vampir GUI remotely via X-forwarding $ vampir & $ vampirserver start -- -P your_project_id Launching VampirServer... [..] User: your_user_name Password: random_password VampirServer <4193> listens on: h41n10:30081
vampir, and connect to the server via Open Other -> Remote File, where you put in the above connection details.