[geos-devel] Thoughts on in/outlining and robustness

strk at refractions.net strk at refractions.net
Wed Jun 14 16:48:02 EDT 2006


Yes, I think you're completely right, much better to work
on the robustness limits. My real concern is facing bogus
results that are not reproducible with JTS.
Luckly this was not the case, but as long as
inlines/optimizations/architectures/compiler_switches
influence the results the problem will always be pending.

Good to know your new approach fixes all the tests we have,
will be interesting to see if GEOS will be as good there.

--strk;

On Wed, Jun 14, 2006 at 12:55:51PM -0700, Martin Davis wrote:
> strk, while I commend your valiant efforts to reconcile the behaviour of
> inlined VS outlined functions and their effect on GEOS robustness, I'm
> wondering whether this effort can ever pay off.
>  
> As I understand the problem, you're worried that inlined VS outlined
> code sometimes produces incorrect results rather than throws exceptions,
> or perhaps even produces correct results.  It makes sense that this is
> due to subtle differences in the way the FP computations are done.  But
> even if you managed to get the two implementations to behave
> identically, there would still be a whole set of test cases which would
> still either get the wrong answer or throw an exception.  JTS has the
> same problem right now.  
>  
> I think the best place to direct effort is in solving the overall
> problem of robustness in the fundamental algorithms.  Hopefully this
> will be solid enough that it will be immune to the inline/outline FP
> differences.  
>  
> My impression is that it would be a lot of incredibly finicky work to
> "solve" the inline/outline difference, without improving robustness all
> that much.  Or am I missing something?
>  
> Martin Davis, Senior Technical Architect
> Vivid Solutions Inc.      www.vividsolutions.com
> Suite #1A-2328 Government Street Victoria, B.C. V8T 5G5
> Phone: (250) 385 6040 - Local 308 Fax: (250) 385 6046
>  

> _______________________________________________
> geos-devel mailing list
> geos-devel at geos.refractions.net
> http://geos.refractions.net/mailman/listinfo/geos-devel


-- 

 /"\    ASCII Ribbon Campaign
 \ /    Respect for low technology.
  X     Keep e-mail messages readable by any computer system.
 / \    Keep it ASCII. 




More information about the geos-devel mailing list