About MSCF

HPC Center Mol Science Software Scientific Consulting Graphics & Vis Lab Partners & Links Research Highlights

Getting Started

Submitting Proposals Accounts Required 1st Time Logging In FAQ

Accounts

Getting An Account Managing Allocations User Support User Policies

Using Resources

File Systems MSCF Software Submit & Manage Jobs Store & Backup Data FAQ

Computer Hardware

Parallel Computers NWfs Graphics & Vis Workstations & Other

Software

Applications NWChem Ecce ParSoft Job Management Compilers System Software

Visualization

VisLab Virtual Tour Services Visualization Gallery Contacts Equipment List

Training

Software Tutorials Workshops/Classes Scientific Consulting User Docs for Apps

Events

Workshops/Classes Meetings

System Reports

System Announcements Current Job Sample NWfs Status User Tips

Click on the | above to Expand/Collapse the menu

•  Home 
•  Site Map 
•  Help 
•  Search 
•  Contacts 
•  Tips 
•  Glossary 
•  Tour 
•  Brochures 

Molecular Science Computing Facility
Computer Hardware

Vampir/Vampirtrace 3.1

Overview

• Vampir is the GUI driven frontend used to visualize the data collected by running a program linked with the vampirtrace library on mpp2.
• Vampirtrace is a library that you link to your object files that collects profile information about MPI jobs on MPP2.

Details

1. Environment Preparation.
   • For csh/tcsh shell users:
     source/home/scicons/apps/Pallas/vampir.csh
   • For sh/bash shell users:
     ./home/scicons/apps/Pallas/vampir.sh

   These scripts set the PAL_ROOT, VT_ROOT, PAL_LICENSEFILE, and VAMPIR_LIB_PATH environment variables, and modify PATH and MANPATH environment variables to add information about the location of the vampir command and its help documentation.

2. Building.

Relink a program with the Vampirtrace library. Start with existing .o files and build the executable by adding:

-L$(VAMPIR_LIB_PATH) -lVT to the end of the link command. For example, if your build commands are:

• mpif90 -c your_prog.f90
• mpif90 your_prog.o -o your_binary

To link with the Vampirtrace library you would use:

• mpif90 your_prog.f90 -o your_binary -L$(VAMPIR_LIB_PATH) -lVT

If you are using HP-MPI (instead of Quadrics MPI), then substitute mpif90.mpich for mpif90, so that the MPICH compatability library is linked in.

3. Data Collection

Execute your program as you normally would (with a bsub), creating the trace file named your_binary.bvt or a large number of trace files (The files generated is software and configuration dependent).

4. Data Analysis

Start Vampir on the .bvt (or .stf) file created in the Data Collection phase.

• vampir your_binary.bvt &

If vampire generated a large number of files. These can be merged into a single stf.single file for easier use (you then run vampir on the stf.single file):

• stftool your_binary.stf --logfile-format STFSINGLE --convert your_binary.stf.single

Depending upon the module environment you are exectuting under, vampir may refuse to run unless a compatibility flag set first:

• export LD_ASSUME_KERNEL=2.4.19 (for bash users)
• setenv LD_ASSUME_KERNEL 2.4.19 (for csh/tcsh users)

Help

Contact the MSCF VisUS Consultants if you have difficulties.

Download current, FREE browsers from Mozilla (any platform), Internet Explorer (Windows only), and Opera (any platform).