I think the semantic description of MPI-IO operations will be very useful.
Some years ago, I came across a ICASE tech report (by Zima, Brezany and
Mahrotra) in which they presented a very similar description. I think the
following tech report will provide some details. (If anyone from Vienna is
reading this mail, please correct me)
@techreport{brezany:cfo,
author = {P. Brezany and M. Gerndt and P. Mehrotra and H. Zima},
institution = {ICASE, NASA, Hampton VA 23681},
month = {September},
number = {ICASE Report 92-46},
title = {Concurrent File Operations in a high Performance Fortran},
year = {1992}
}
rajesh
------------------------------------------------------------------------------
Rajesh R. Bordawekar Center for Advanced Computing Research
158-79, California Institute of Technology, Pasadena, CA 91125
rajesh@cacr.caltech.edu http://www.cacr.caltech.edu/~rajesh
office: 818-395-2909 fax : 818-584-5917
------------------------------------------------------------------------------