<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">Ghosh:<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="font-family:'times new roman','new york',times,serif;font-size:12pt;color:rgb(0,0,0)"><div><br></div><div> Thanks for the suggestion of Elemental. However I need to have the dense matrix have the same parallel communicator as the sparse matrix. </div><div>If I use elemental, then the numbering changes, will I be able to multiply with the sparse matrix which has a different numbering scheme and communicator?</div><div><br></div><div>In any case I would want the resultant matrices A and M to be elemental type since I need to solve a dense eigenvalue problem.</div></div></blockquote><div> </div><div>If your final goal is solving a <span style="color:rgb(0,0,0);font-family:'times new roman','new york',times,serif;font-size:16px">dense eigenvalue problem, and the only sparse matrix is a banded matrix H, Elemental is a suitable package. </span></div><div><span style="color:rgb(0,0,0);font-family:'times new roman','new york',times,serif;font-size:16px"><br></span></div><div><span style="color:rgb(0,0,0);font-family:'times new roman','new york',times,serif;font-size:16px">I just added </span><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px">HermitianGenDefiniteEig() into petsc(master)-elemental(v0.84) interface and tested on some large matrices - seems quite efficient. </span></font><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px">You may take a look at ~</span></font><span style="color:rgb(0,0,0);font-family:'times new roman','new york',times,serif;font-size:16px">petsc(master)/</span><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px">src/mat/examples/tests/ex174.cxx</span></font></div><div><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px"><br></span></font></div><div><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px">I'm not sure if the latest Elemental supports banded dense matrix - if not, ask Jack to added :-)</span></font></div><div><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px"><br></span></font></div><div><font color="#000000" face="times new roman, new york, times, serif"><span style="font-size:16px">Hong</span></font></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="font-family:'times new roman','new york',times,serif;font-size:12pt;color:rgb(0,0,0)"><div><br></div><div><hr><div style="color:rgb(0,0,0);font-weight:normal;font-style:normal;text-decoration:none;font-family:Helvetica,Arial,sans-serif;font-size:12pt"><b>From: </b>"Hong" <<a href="mailto:hzhang@mcs.anl.gov" target="_blank">hzhang@mcs.anl.gov</a>><br><b>To: </b>"Swarnava Ghosh" <<a href="mailto:sghosh2012@gatech.edu" target="_blank">sghosh2012@gatech.edu</a>><br><b>Cc: </b>"PETSc users list" <<a href="mailto:petsc-users@mcs.anl.gov" target="_blank">petsc-users@mcs.anl.gov</a>><br><b>Sent: </b>Thursday, February 5, 2015 8:22:13 PM<br><b>Subject: </b>Re: [petsc-users] Large rectangular Dense Transpose multiplication with sparse<div><div class="h5"><br><br><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">Swarnava:</div><div class="gmail_quote">The matrix product A will be a dense matrix. You may consider using Elemental package for such matrix product.</div><div class="gmail_quote"><br></div><div class="gmail_quote">Hong<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><br>
Dear all,<br>
<br>
I am trying to compute matrices A = transpose(R)*H*R and M = transpose(R)*R where<br>
H is a sparse (banded) matrix in MATMPIAIJ format (5 million x 5 million total size)<br>
R is a MPI dense matrix of size 5 million x 2000.<br>
<br>
I tried 1) MatPtAP - Failed, realized this only works for pairs of AIJ matrices<br>
2) First multiplying H*R and storing in 5 million x 2000 MPI dense. Then MatTranspose of R and multiplying the transposed R with 5 million x 2000 dense. This multiplication fails.<br>
<br>
Could someone please suggest a way of doing this.<br>
<br>
Regards,<br>
Swarnava<br>
<span><font color="#888888">--<br>
Swarnava Ghosh<br>
</font></span></blockquote></div><br></div></div>
</div></div></div><span class=""><font color="#888888"><br><br><br>-- <br><div><span name="x"></span>Swarnava Ghosh<br>PhD Candidate,<br>Structural Engineering, Mechanics and Materials<br>School of Civil and Environmental Engineering<br>Georgia Institute of Technology<br>Atlanta, GA 30332<span name="x"></span><br></div></font></span></div></div></blockquote></div><br></div></div>