Re: [Gomp-discuss] GOMP Requirements v1.1

[Ross Towle]

On Fri, 12 Nov 2004 09:35:50 -0500, Scott Robert Ladd
<address@hidden> wrote:

> Ross' document is a good piece of work, but it is not a complete design
> document. We have several issues that need to be seriously considered
> before we know all the details of hos this is going to work. In many
> ways, I think we've gotten that cart before the horse; we've implemented
> a support library with questionable copyright legalities, and we have
> designs before we even define what all the requirements are.

I agree that my document is not complete.  It was not intended to be
complete.  I wanted to present a potential direction for
implementation.  If that direction was acceptable to the group I would
work on additional detail.  I was worried if I initailly presented too
much detail then the general idea would be lost in those details.

> 
> The real question is: Does GCC care about being competitive in terms of
> performance? If intellectual freedom is the only goal, then we can by
> all means approach this from a generic perspective. If we want a
> compiler than is a practical alternative to commercial products, we need
> to concern ourselves with performance issues.

The generic approach does not preclude OpenMP performance.  I think if
you would examine the commercial implementations, they go for a
generic approach. They hide the details of using LWP (lightweight
processes), SPROCS, threads, or whatever threading model in the
support routines.  The user may have no choice of the threading model
but the threading model is not in the generated user's code.  It is in
the library.

The commercial implementations also use the generic approach to assist
correctness and to make it easier to follow the changes to the OpenMP
specification from 1.0 to 2.0 to 2.5 to 3.0 to ... without a major
rewrite or revisit to the design.  The changes to the spec work as
extensions.

Cheers - Ross A Towle