[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Tue, 6 Apr 2004 10:48:06 +0200
I have added Jacob to the developers of the project. He plans to do
housekeeping following what he has exposed in his TODO (which by the way
makes a lot of sense to me).
Jacob's letter gave me also another opportunity to think why we haven't done
so much 'till now. One obvious reason is lack of personal time by almost all
of us, and there is nothing one can do about it. But in my opinion there are
other reasons easier to deal with. The problem so far is that we have been
planning to write a very genera framework for parallelisation, but almost
no-one knows how to do it. So, let's fly lower and redefine our target.
I suggest we focus our efforts on Linux, defining our objectives step by step.
If there is some interest for some other architecture, someone will step in
and contribute. Obviously the risk of not being generic enough is that you
might have to go through several rewriting. I am prepared to pay this price,
if this will lead to a correct parsing of hello_omp.c in a relatively short
As you have said countless times, the obvious areas of intervention are:
I would leave aside (c) for a moment and resolve possible conflicts in the
optimisation step by imposing the -O0 flag to every compiled program.
The development of the lib can be done independently of anything else, while
the development of the frontends can build on the stubs. Without going too
much ahead, I propose that we endorse the 0.1 plan proposed by Jacob for the
frontends, while complete freedom is left for the moment on the libs.
Since we have to play with our own strengths, we don't want to build the most
elegant implementation around nor want we to have the most efficient. Let's
have something that Just Works (TM).
Opinion, corrections and insults in this list, please :-)
Institut Fuer Theoretische Physik
CH-8093 Zuerich - Switzerland
Tel. +41 (0)1 6332562
|[Prev in Thread]
||[Next in Thread]|
- [Gomp-discuss] Housekeeping,
Biagio Lucini <=