[boost] [Review]{MPI] boostMPI review

From: John Phillips (phillips_at_[hidden])
Date: 2006-09-12 10:07:55

- What is your evaluation of the design?

   I like the choice of making it focus on easing the communications 
tasks using the serialization library. At least for those new to MPI, 
proper message construction and interpretation is a major hurdle that 
this greatly reduces.

- What is your evaluation of the implementation?

   Seems solid, but I didn't get very deep into it.

- What is your evaluation of the documentation?

   In need of editing for detail. Numerous typos and some questionable 
sentence constructions.
   Also useful would be adding a translation table. For those new to the 
MPI, a well organized listing of C style MPI functions and the boostMPI 
equivalent would allow them to translate example code in C to similar 
code in boostMPI bindings. Since there are a large number of books and 
online references listing usable code in C and Fortran, but none yet use 
boost MPI thas would speed adoption of the library.

- What is your evaluation of the potential usefulness of the library?

   I think it will be very useful. As it stands, it can support many 
reasonable MPI based tasks, and when it has matured to support all of 
MPI 2, it will be the obvious choice for MPI programming in C++.

- Did you try to use the library? With what compiler?
    Did you have any problems?

   I did a few simple examples using gcc compilers with a 
multi-processor Mac box. I also did them on a small cluster of 
Windows/Cygwin boxes and on a small cluster of Sun boxes. The underlying 
MPI implementation was OpenMPI for the mac and for the Windows boxes, 
and LAM/MPI for the Sun boxes. I didn't have any problems.

- How much effort did you put into your evaluation? A glance? A
    quick reading? In-depth study?

   A few hours, so not all that long. I read the docs fairly closely, 
installed the libraries and wrote, compiled and ran some example code. I 
have not yet used it in reasearch code, but I am likely to do so in the 
next few months.

- Are you knowledgeable about the problem domain?

   I'm somewhat knowledgeable. I make use of MPI for research code and I 
am currently teaching a class on parallel and distributed programming 
using MPI. I have not used it in industrial settings.

   The bottom line for me is a request for additions to the 
documentation, a hope that full MPI 2 support can become a reality soon, 
and a recomendation that the library be included in boost.

                        John Phillips