The tutorial itself covers examples from chapters 3 through 5 of ``Using MPI...''. I also provide detailed references to the MPI definition document about which more below. I have tweaked the examples so as to get them going on the ``LAM engine''. Some examples had to be modified more than others. In particular Argonne MPICH assumes that the console process always has rank 0. But this is not specified in the MPI standard, and LAM does it differently. This has an impact on how you do I/O. These issues have been left out of the MPI definition because they are implementation and vendor dependent. For example on a workstation farm you can always rely on standard UNIX I/O. Even if you cannot write directly to your VDU, you can always open a file in /tmp or /usr/tmp and write there. But this may not be possible if you run your MPI programs on a machine which does not support full UNIX on every node.
Lack of I/O specification is a shortcoming of MPI library, acknowledged in ``Using MPI...'' (section 10.4, page 245) and in the MPI definition document (section 1.5, page 3). But this shortcoming is at present insurmountable due to the fact that the idea of installing a full UNIX I/O on every node has not been universally accepted by all supercomputer vendors.
Another specific feature of this tutorial is that it is restricted to pure MPI. You will find quite a few references to an extension package available with MPICH in ``Using MPI...''. Those extensions, which are enclosed in the MPE library, are not necessarily supported by other MPI implementations. LAM specific features, such as using the LAM frame buffer and LAM LED facility will be covered in the companion to this tutorial later.