Re: [Gomp-discuss] OpenMP 3.0 Documents?

[Lars Segerlund]

 I think we'll be safe as long as we make the 'middle end' , ( by this I mean 
tree-ssa extensions, ( yes GENERIC too ) ), the backend ( the lib ) as generic 
as we can, in other words keep them simple and only add the minimum set 
required to handle the task at hand. The more we make them specific the harder 
everything will get in the future, ( i.e. if someone else want's to add HPF or 
something similar ).
 A lot of the hard part must be to reduce the complex task into a minimal 
functionality that is required, something must already be done in this context, 
so I will google a bit for it.

 The part that we absolutely must get right is the representation of 
parallellism in tree-ssa, ( it must be independent of the frontend used ). Thus 
all the semantics and I must admit a big deal of the complexities are moved 
into the frontends.

 Is this correct thinking or plain dumb ?

 / Lars Segerlund.


On Wed, 3 Nov 2004 16:37:26 +0200
Biagio Lucini <address@hidden> wrote:

> On Wednesday 03 November 2004 13.51, Scott Robert Ladd wrote:
> >
> > However, I'm not a registered member of the OpenMP group (it costs
> > $8000). Does anyone else have these documents?
> >
> 
> Registered members are usually big corporations, I think you have little 
> chances to find one among us. I suggest you go on, the draft should be public 
> in some months, even before it gets final.
> 
> Cheers
> 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