Hi,

On Fri, Nov 14, 2014 at 04:33:45PM +0100, Bård Skaflestad wrote:
> On 2014-11-14 15:33, Jørgen Kvalsvik wrote:

> >I am able to reduce the running time of
> >upscale-benchmark-relperm. I attribute this to a more efficient
> >allocation and instantiation of the matrices, as the BCRSMatrix now can
> >be build row-wise and not in the more inefficient random mode.
> 
> Do keep in mind, however, that per-row construction leads to a data
> structure in which the matrix values are stored in separate
> containers per row whereas the original scheme uses a single
> contiguous array for the values.  That may or may not matter in any
> particular application.
> 

If you provide the number of nonzeros to the constructor, then  using
when using row-wise mode there will be only one container for row start,
column index and values. 

Markus

-- 
Dr. Markus Blatt - HPC-Simulation-Software & Services http://www.dr-blatt.de
Hans-Bunte-Str. 8-10, 69123 Heidelberg, Germany,  USt-Id: DE279960836
Tel.: +49 (0) 160 97590858

Attachment: signature.asc
Description: Digital signature

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

Reply via email to