Parallel Programming Guide for HP-UX Systems

MPI

Tuning

Chapter 2 67

In this case, all iterations through MPI_Recv_init are progressed just once when

MPI_Startall is called. This approach avoids the additional progression overhead when

using MPI_Irecv and can reduce application latency.

Multiple network interfaces

You can use multiple network interfaces for interhost communication while still having

intrahost exchanges. In this case, the intrahost exchanges use shared memory between

processes mapped to different same-host IP addresses.

To use multiple network interfaces, you must specify which MPI processes are associated with

each IP address in your appﬁle.

For example, when you have two hosts, host0 and host1, each communicating using two

ethernet cards, ethernet0 and ethernet1, you have four host names as follows:

• host0-ethernet0

• host0-ethernet1

• host1-ethernet0

• host1-ethernet1

If your executable is called beavis.exe and uses 64 processes, your appﬁle should contain the

following entries:

-h host0-ethernet0 -np 16 beavis.exe

-h host0-ethernet1 -np 16 beavis.exe

-h host1-ethernet0 -np 16 beavis.exe

-h host1-ethernet1 -np 16 beavis.exe

Now, when the appﬁle is run, 32 processes run on host0 and 32 processes run on host1 as

shown in Figure 2-2.