pj@linz.ensmp.fr (Pierre JOUVELOT) (01/11/91)
We are interested in improving our programming environment, targeted
towards the interactive parallelization of Fortran programs. The picture
is the following. Each analysis used by our parallelizer builds new data
structures based upon other information. For instance, the call graph of
a whole program is computed after the determination of the symbol table
and the code of each module. Now, we want to avoid useless
recomputations in case some of these information get changed by the
user. In our current system, dependencies between these information are
declaratively defined and each phase consumes and produces some
information.
We are aware of the work done in the area of program version management,
but are wondering whether other bibliographic sources are available that
deal with this sort of question.
The kind of issue we are worried about is the seemingly combinatorially
explosive process that arises when options, used in some phases, can be
changed at will by the programmer (e.g., to select a more stringent
dependence test). In that case, the whole history that led to the
computation of an information seems to have to be kept around ... which
seems kind of expensive :-)
Please send your remarks or comments to <baron@ensmp.fr>.
Pierre
--
Pierre Jouvelot
. CAII, Ecole des Mines de Paris, 35 rue Saint Honore', 77305 Fontainebleau,
France (jouvelot@ensmp.fr)
. LCS, MIT, USA (jouvelot@lcs.mit.edu)
--
Send compilers articles to compilers@iecc.cambridge.ma.us or
{ima | spdcc | world}!iecc!compilers. Meta-mail to compilers-request.