<html><head><style type="text/css"><!-- DIV {margin:0px;} --></style></head><body><div style="font-family:times new roman, new york, times, serif;font-size:12pt"><div><jinzishuai@yahoo.com>> Then my answer is the same as my second answer. To first order, I think the time<br>> is completely negligible, so just use an AIJ matrix. If, after measuring the<br>> performance, you are unhappy, then put the diagonal matrix in a vector and<br>> wrap MatDiagonalScale() in a MatShell.<br>> <br>> Matt<br>Thank you. I think I am OK with AIJ matrix and it is working fine.<br>But I still want to ask the first question about specifying a preconditioning matrix. Right now I am testing a first order element case since I know what it should look like and it should improve performance. But when I do <br><jinzishuai@yahoo.com><br>> > > ierr = KSPSetOperators( solMP, M, lumpedM, SAME_PRECONDITIONER);<br><br>The solver takes
more iterations to converge. It this the correct way to specify the lumped matrix?<br>Thank you very much.<br><br>Shi<br><br>> > ><br>> > > However, instead of giving faster convergence, it actually takes more<br>> > > iterations to convergence than the regular one. Therefore, I wonder if<br>> > > setting lumpedM as Pmat is the correct way to do it. Could you please<br>> > > advice? I think right now lumpedM is taken as the input to compute the<br>> > > preconditioning matrix, using whatever method is specified by -pc_type .<br>> > > What I really want to do is to simply set lumpedM as the precondition<br>> > > matrix, without spending time to compute anything.<br>> > > Thank you very much.<br></jinzishuai@yahoo.com></jinzishuai@yahoo.com></div><br></div><br>
<hr size=1>OMG, Sweet deal for Yahoo! users/friends: <a href="http://us.rd.yahoo.com/evt=47520/*http://tc.deals.yahoo.com/tc/blockbuster/text2.com
">Get A Month of Blockbuster Total Access, No Cost.</a> W00t</body></html>