awips2/pythonPackages/scientific/README.BSP

BSP programs in Python can be executed in three modes:

1) Using a standard Python interpreter. No special measures need to
   be taken, the program will execute on a single processor.

2) Using MPI or BSPlib for communication between multiple CPUs.
   This is the only mode that gives real parallel execution. See
   below for details.

3) Using the bsp_virtual interpreter.
   This runs the program with any number of virtual processors,
   but only one real CPU is used. The main interest of this
   mode is for debugging, as it permits interactive debugging
   of any or all virtual processors. See below for details.

   Note: this code is not very well tested yet. In particular,
   it might fail with programs that manipulate sys.modules
   directly.


Real parallel execution
-----------------------

In order to use the module Scientific.BSP using more than one real
processor, you must compile either the BSPlib or the MPI interface.
See README.BSPlib and README.MPI for installation details. The BSPlib
interface is probably more efficient (I haven't done extensive tests
yet), and allows the use of the BSP toolset, on the other hand MPI is
more widely available and might thus already be installed on your
machine. For serious use, you should probably install both and make
comparisons for your own applications. Application programs do not
have to be modified to switch between MPI and BSPlib, only the method
to run the program on a multiprocessor machine must be adapted.

To execute a program in parallel mode, use the mpipython or bsppython
executable. The manual for your MPI or BSPlib installation will tell
you how to define the number of processors.

Interactive execution in parallel mode is possible using the scripts
ibsppython and impipython. Note that the latter must be adapted to
your MPI installation. These scripts emulate a standard Python
interpreter well enough that they can be used in some developmen
environments. In particular, they can be used within Emacs Python
mode.


Virtual parallel execution
--------------------------

The script bsp_virtual implements a BSP virtual machine that
serializes the execution of any number of virtual processors. There is
no significant overhead, but there isn't any real parallelism either,
only one CPU is used. However, this mode allows the most convenient
interactive development, including interactive debugging on all
virtual processors. In combination with Emacs and its Python mode, it
is a very convenient development environment for parallel programs.

bsp_virtual emulates a standard Python interpreter even better than
the interactive BSP scripts mentioned above. In addition, it provides
interactive debugging of virtual processes. The commandline syntax is

   bsp_virtual np [options] [script.py]

where np is the number of (virtual) processors and script.py is
a Python script to execute. The options are

   -i:  continue in interactive mode after the script terminates
   -r:  print the runtime used by each virtual processor

Debugging is handled by the standard pdb, which however is invoked in
a different way. Debugging commands are identified by an exclamation
mark (!) at the beginning of the command line. The following two
debugging commands are available:

   !pm(pid)
      Runs a post-mortem inspection after an exception. pid is the
      processor id to be inspected. Note that *all* processor states
      can be inspected, not only the one of the process that crashed.

   !run(pids, command)
      Run command (a string) under debugger control. pids is a list
      of processor ids that are subject to debugging. It can also be
      an integer instead of a list, for debugging a single process.
      The debugging commands are the same as in the standard pdb.