Re: [Gomp-discuss] The tree nodes ...

[Steven Bosscher]

Op ma 10-02-2003, om 15:05 schreef Lars Segerlund:
> 
>   I have been reading Diego's paper this weekend, and It seem's like 
> just the starting point we need, however we might want another set of 
> tree nodes, as there are a couple of explicit attributes which openMP 
> specifies which was not covered in their work.
> 
>   ( such as ATOMIC variables )
> 
>   Is there any toughts on the primitives needed, and how we should go 
> about specifying the implementation we aim for ?
> 
>   I think that as soon as we have the tree form specified we can start 
> work on all different parts, the parsing into the tree's, the 
> implementation of the same, and the integration into the optimization 
> and code generation routines.

Sounds good.  Let's first try to define the primitives we need: What is
the absolute mimimum set of operations needed for OpenMP (e.g. which
pragmas can be expressed as a sequence or set of the others? In other
words, which ones are convenient but actually redundant?).
 
>   ( also want to vent my opinion that we should make the distinction 
> between section running on the same machine and sections slated to 
> target another cpu ).

Mwah... I'm not sure that's a good idea.  I doubt you can actually
control which CPU a thread will run on.  Can you?

>   As for the discussion about HPC gnu initiative, I think that we can 
> postpone that and rather work with the existing standards, such as 
> Threading libraries and open MPI or PVM implementations, ( the last two 
> in the future ), but what it actualizes is the need not to restrict 
> ourself to much.

Agree.  First, define explicit concurrentcy in GCC.  Then do OpenMP with
that.  We can see what happens when that's finished.

Greetz
Steven