Re: OpenMP and its testing

[Diego Novillo <dnovillo at redhat dot com>]

On Wed, 2004-11-10 at 07:15, Nick Maclaren wrote:
> Frankly, I think that this is asking for trouble.  We are a parallel
> shop, and my experience is that most C OpenMP implementations are
> very sick - my experience is that only half of the major vendors'
> implementations work at all.  The reasons should be well-known, but
> please ask me if anyone wants to know more.
> 
> Anyway, I have some crude tests that I wrote because I couldn't find
> any on the net.  These are based on versions of C and Fortran Linpack
> and Schwarztrauber's FFT that I got off some site (perhaps netlib).
> I am happy to allow free use of any parts that are my copyright,
> but the majority is of course the original authors'.  And there don't
> seem to be any copyright notices attached, but I am prepared to state
> that I have not removed any and that my understanding was that they
> were usable without conditions.
> 
> If you are interested, I think that it would be fairly easy to check
> up on that.
> 
Yes, that sounds great.  Even if you are not in a position of releasing
the copyright to the FSF, we could still work out a way of either having
you running regular tests or making the testsuite available for other
GCC developers to test.

> Anyway, even if not, I am very happy to collaborate on the testing (and
> perhaps even development) of OpenMP application-style tests and timers.
> I can test them on a range of other vendors' OpenMP compilers.  I am
> NOT a CVS user and hope to remain that way!
> 
Unless you started contributing to development on a regular basis, you
wouldn't need to have CVS access.  Regular snapshots are always
available.


Thanks.  Diego.