Re: [Gomp-discuss] A suggestion.

[Lars Segerlund]

 I have harvested citeseer for some usefull information and came up with a 
bunch of papers on OpenMP which I'm going through right now.

 I think it would be preferable if we could put these somewhere publicly 
available ( or atleast the ones that we reference or use ), ( an alternative is 
to put links to them on citeseer ).

 What do you think ?

 There is a lot of stuff on the SSA form and on the runtime systems available 
and also NUMA/MPI comparisons, mixes and so.

 A lot of the discussions here will be a lot easier if we have a document 
archive on our webpage or something similar, since direct references are always 
easier than having to google for your self.

 I will try to get a list of the most relevant ones by early next week.

 What I'm mostly concerned about is that we have agree on some kind of subset 
or representation of functionality to bring through all the interfaces, ( ie. 
primitives ). These have to be expressive enough to be able to handle the 
semantics of C,C++ and fortran in the 'middle end'.
 I know we have talked a lot about the differences, but  we don't have to hard 
code them if we have a reasonable representation to bring into the backend.

 / Lars Segerlund.


On Mon, 08 Nov 2004 09:13:47 -0500
Scott Robert Ladd <address@hidden> wrote:

> Lars Segerlund wrote:
> > I have the suggestion that for the gomp project nothing goes in witch
> > is not documented or referenced, as an absolute requirement.
> 
> I agree. I'm trying to coallesce all of the exiting design documentation 
> into a comprehensive document; I hope to get the preliminary version out 
> this week.
> 
> ..Scott
> 
> -- 
> Scott Robert Ladd
> site: http://www.coyotegulch.com
> blog: http://chaoticcoyote.blogspot.com
> 
> 
> _______________________________________________
> Gomp-discuss mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gomp-discuss