On Thu, 2013-10-10 at 13:51 +0200, Roland Kaufmann wrote:
> On 2013-10-10 13:45, Bård Skaflestad wrote:
> > do recall that someone (Markus?) pointed out on that at least on
> > supercomputers, the (startup) cost of the dynamic linker resolving
> > symbols is considerable.
> 
> I don't know if it should be classified as considerable, but the
>  dynamic linking has to be done on every node, so it has zero
>  scalability. (This is where supercomputer people love to refer to
>  Amdahl's "law").

The numbers that were quoted were in the order of 30 minutes on a large
application and I'd say that's "considerable".  Unfortunately, I can't
find the reference right now so the point is moot and I won't pursue the
issue any further.  Maybe it was 30 minutes total CPU time aggregated
across all nodes rather than wall time.


Bård


_______________________________________________
Opm mailing list
[email protected]
http://www.opm-project.org/mailman/listinfo/opm

Reply via email to