So for dense matrices, we assume column alignment when we pass in the pointer? That is probably what was wrong. Crap it does say this at the top, but not at the argument. I will augment the documentation.
Matt On Sun, Jan 17, 2010 at 10:42 AM, Jed Brown <jed at 59a2.org> wrote: > On Fri, 15 Jan 2010 09:58:46 -0600, Matthew Knepley <knepley at gmail.com> > wrote: > > SeqDense MatMult was inexplicably giving me the wrong answers. I > > checked the BLAS call (and the underlying defines) and they also > > looked right to me. I temporarily checked in the simple-minded loops > > which do give me the correct answer. > > The matrices are column-aligned. The "simple-minded loops" you checked > in (cc0b86d7fd3f) assume row-aligned matrices, and are thus delivering > the wrong results for everyone that assembled their matrix correctly, so > I'm reverting that change. > > How were you assembling this matrix? > > Jed > -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20100117/343a00e5/attachment.html>