Re: [Gomp-discuss] Release candidate

[Lars Segerlund]

 Get it in there and get some feedback :-) that's my opinion.

 OpenMP is alredy mentioned on the gfortran pages, so I think were already made 
a bit of a stir.

 I disagree slightly about the libs and their design, ( based on looking at 
intels openMP compiler which handles threads a bit different in order to get 
higher performance on the numa parts, and my wishes to be able to run on MPI or 
something similar, there are a lot of clusters showing up nowadays and I think 
we need so supprot them ). ( Ok another shameless plug for my humble opinions 
:-) ).

 I'm doing some stuff in gfortran which indirectly relates to openMP in that 
it's parser and backend related, and have been doing some small tests on the 
fortran frontend and it's handling of comments in order to see if the interface 
I have grafted onto gfortran also can be used for the openmp directives, so far 
it looks ok, I have a couple of show stoppers :-) .. ( nothing bad but ! ).

 The approach I have is to hack a generic plugin for directives in comments, 
and then all our code would be selfcontained, this is fairly easy to do with 
the fortran parser so I hope it will pay off.
 ( Getting the generic interface accepted into gfortran and then we can hack on 
on our own ).

 I think we shall keep as many options open as possible while remembering the 
words of Linux Thorvalds, never overdesign, get something running first, it's 
more important. If we need changes later we can just rip it apart and stick it 
together again.

 / Lars Segerlund.

On Tue, 2 Nov 2004 14:13:08 +0200
Biagio Lucini <address@hidden> wrote:

> Hi Jacob, thanks for your comments.
> 
> On Tuesday 02 November 2004 12.55, Jacob Weismann Poulsen wrote:
> >
> > 2) As to the content of libgomp I made a few early experiments
> > that wasn't meant to make its way to the official tree. The
> > reason is simply that I didn't really think about the structures
> > nor the implementation and as Scott has said on several occasions
> > (and I think that we all agree on this), we better base the
> > implementation on some kind of design document. I would recommend
> > that the initial commit will take its offspring in the contents
> > at the savannah CVS instead or in version 1.1 of the files in the
> > cvs-ball that I mailed you and Scott a long time ago (if I recall
> > it right - it seems to be ages ago).
> >
> 
> Well, this has been debated quite a lot. I was sharing your opinion at the 
> beginning, but Diego sort of convinced me that this is not the case, since
> having gomp into some branch of gcc will help us to get visibility and to 
> receive more comments from other people and get more people involved too. 
> Putting out some code at this stage could also allow us to receive 
> constructive feedback for the design. If you are concerned of going into the 
> official tree, my understanding is that a branch is not an official tree and 
> you are free of doing on it whatever you like (subject to the maintainer 
> approval, of course). All those arguments to me seem enough to submit a patch 
> to gcc-patches for the gomp_branch. Let me stress that with this patch you 
> can compile and run OpenMP program (albeit in dummy mode): what this means is 
> that people will get for the first time a proof of the existence of OpenMP 
> within gcc. Then, maybe everything will need to be erased after we have a 
> good design (which - I agree with you - is fundamental). But meanwhile we 
> would have discussed, we would have received input, we would have received 
> more help and perhaps even flames that can be useful for future development.
> 
> Also, for my personal point of view, I have learnt a lot by working just on 
> the building of the tree within gcc. This would not have been possible by 
> simply committing the first tarball into savannah CVS. I think there is lots 
> more than can be learnt by using a pragmatic approach. However, this is 
> mainly your code, and my opinion here counts very little; but since the stubs 
> are public, I would like at least to submit a libgomp_stubs. So let me know 
> if you want some code to be removed from the current tree, and I will take 
> fully responsibility myself for the rest. Of course, similar remarks also 
> apply to the other two of you that have contributed code, namely Scott and 
> Lars. 
> 
> Thanks,
> Biagio
> 
> -- 
> =========================================================
> 
> Biagio Lucini                               
> Institut Fuer Theoretische Physik
> ETH Hoenggerberg      
> CH-8093 Zuerich - Switzerland           
> Tel. +41 (0)1 6332562  
>  
> =========================================================
> 
> 
> _______________________________________________
> Gomp-discuss mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gomp-discuss