Starting HPF Programs

The way in which TotalView starts an HPF parallel program depends on the machine on which the code is running and the runtime library linked into the HPF code.

PGI HPF smp and rpm Libraries

Using TotalView to start a program linked with the smp and rpm libraries is similar to the way in which you would normally start the program. For example, suppose you start the program as follows:

my_program -bah -pghpf -np 6

Here is the command you would use to debug this file:

totalview my_program -a -bah -pghpf -np 6

Starting HPF Programs with MPICH

In a workstation cluster environment using MPICH, debug your HPF application with TotalView by adding the -tv option to the mpirun command. For example, assume that you would begin executing your code with the following command:

mpirun -np 4 my_program

Using mpirun, you would invoke TotalView as follows:

mpirun -tv -np 4 my_program

Workstation Clusters Using MPICH

Debugging workstation clusters uses the same mechanism as debugging an MPICH program since a compiled HPF program is an MPICH program. For more information, refer to Debugging MPI Applications Overview.

IBM Parallel Environment

In the IBM parallel environment on an IBM SP or cluster of RS/6000 machines, parallel programs are started by using the poe command. Here is an example of starting TotalView on the poe command to debug a program named hpf_test:

totalview poe -a hpf_test -procs 6

For more information, refer to Starting TotalView on a PE Job.