[Trilinos-Users] LOBPCG direct diagonalization
임재창
ljchang94 at kaist.ac.kr
Mon May 9 07:04:33 EDT 2016
Dear Trilinos developers and users,I'm using Anasazi package to diagonalize sparse matrices with Epetra_CrsMatrix.The matrix dimension is about 1 million and i want to get about 1200 number of lowest eigenvalues and eigen vectorsThe elpased time of each part is as followas you can see "Direct solve" takes much more time than other processes. I guess this part is not parallelized.Anyone knows how to accelerate this part or perform in parrallel manner?thank youTimer Name MinOverProcs MeanOverProcs MaxOverProcs MeanOverCallCounts------------------------------------------------------------------------------------------------------------------Anasazi: LOBPCG::Computing residuals 4.096 (12) 4.339 (12) 4.538 (12) 0.3616 (12)Anasazi: LOBPCG::Direct solve 910.2 (12) 925.8 (12) 1024 (12) 77.15 (12)Anasazi: LOBPCG::Initialization 70.62 (4) 70.65 (4) 70.78 (4) 17.66 (4)Anasazi: LOBPCG::Local projection 63.85 (12) 188.2 (12) 308.4 (12) 15.68 (12)Anasazi: LOBPCG::Local update 35.46 (12) 133.4 (12) 149.1 (12) 11.12 (12)Anasazi: LOBPCG::Operation M*x 0 (0) 0 (0) 0 (0) 0 (0)Anasazi: LOBPCG::Operation Op*x 10.63 (12) 130.9 (12) 255.2 (12) 10.9 (12)Anasazi: LOBPCG::Operation Prec*x 0 (0) 0 (0) 0 (0) 0 (0)Anasazi: LOBPCG::Orthogonalization 96.28 (12) 96.61 (12) 97.86 (12) 8.051 (12)Anasazi: LOBPCG::Sorting eigenvalues 0.8634 (12) 1.014 (12) 1.4 (12) 0.08448 (12)------------------------------------------------------------------------------------------------------------------Anasazi: LOBPCGSolMgr locking 0 (0) 0 (0) 0 (0) 0 (0)Anasazi: LOBPCGSolMgr::solve() 1482 (2) 1482 (2) 1482 (2) 741 (2)Epetra_CrsMatrix::Multiply(TransA,X,Y) 11.22 (14) 131.7 (14) 256.1 (14) 9.407 (14)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://trilinos.org/pipermail/trilinos-users/attachments/20160509/8bcdc47b/attachment.html>
More information about the Trilinos-Users
mailing list