On Mar 24, 2012, at 2:33 PM, Roy Stogner wrote:
>
> On Sat, 24 Mar 2012, Derek Gaston wrote:
>
>> Yep - we've known about this for a while. We have special
>> preallocation routines that spew zeros into the matrix for stuff like
>> future contact interactions before the first matassemblybegin/end.
>
> That's a clever but appalling hack. Is there no way to disable the
> sparsity compression with a PETSc call or command line flag?
No, the "sparsity compression" is Intrinsic to the AIJ, BAIJ, and SBAIJ Mat
implementations. Without it you have array access "with many holes" that
effects performance (and complicates codes).
It would be way cool if someone wrote another Mat implementation that was
more dynamic in this regard, but it would be a lot of work and would require
careful design to get the same performance as the current implementations. Of
course, as always the user interface would be the same.
Barry
> ---
> Roy
>
> ------------------------------------------------------------------------------
> This SF email is sponsosred by:
> Try Windows Azure free for 90 days Click Here
> http://p.sf.net/sfu/sfd2d-msazure
> _______________________________________________
> Libmesh-devel mailing list
> Libmesh-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/libmesh-devel
------------------------------------------------------------------------------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here
http://p.sf.net/sfu/sfd2d-msazure
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
