I am trying to solve a complex matrix equation which was assembled using MatCompositeMerge using MUMPS and LU preconditioner. It seems to me that the solve is stuck in the factorization phase. It is taking 20 mins or so, using 16 processes. A problem of the same size using reals instead of complex was solved previously in approximately a minute using 4 processes. Mumps output of <i>-mat_mumps_icntl_4 1 </i>at the end of this email. Does anyone have any ideas about what the problem maybe ?<div>
<br></div><div>Thanks, </div><div><br></div><div>-Nachiket<br><div><i><br></i></div><div><i><br></i><div><br></div><div><div>Entering ZMUMPS driver with JOB, N, NZ = 1 122370 0</div><div><br></div><div>
ZMUMPS 4.10.0 </div><div>L U Solver for unsymmetric matrices</div><div>Type of parallelism: Working host</div><div><br></div><div> ****** ANALYSIS STEP ********</div><div><br></div><div> ** Max-trans not allowed because matrix is distributed</div>
<div> ... Structural symmetry (in percent)= 100</div><div> Density: NBdense, Average, Median = 0 42 26</div><div> Ordering based on METIS </div><div> A root of estimated size 2736 has been selected for Scalapack.</div>
<div><br></div><div>Leaving analysis phase with ...</div><div>INFOG(1) = 0</div><div>INFOG(2) = 0</div><div> -- (20) Number of entries in factors (estim.) = 563723522</div>
<div> -- (3) Storage of factors (REAL, estimated) = 565185337</div><div> -- (4) Storage of factors (INT , estimated) = 3537003</div><div> -- (5) Maximum frontal size (estimated) = 15239</div>
<div> -- (6) Number of nodes in the tree = 7914</div><div> -- (32) Type of analysis effectively used = 1</div><div> -- (7) Ordering option effectively used = 5</div>
<div>ICNTL(6) Maximum transversal option = 0</div><div>ICNTL(7) Pivot order option = 7</div><div>Percentage of memory relaxation (effective) = 35</div>
<div>Number of level 2 nodes = 35</div><div>Number of split nodes = 8</div><div>RINFOG(1) Operations during elimination (estim)= 4.877D+12</div>
<div>Distributed matrix entry format (ICNTL(18)) = 3</div><div> ** Rank of proc needing largest memory in IC facto : 0</div><div> ** Estimated corresponding MBYTES for IC facto : 3661</div>
<div> ** Estimated avg. MBYTES per work. proc at facto (IC) : 2018</div><div> ** TOTAL space in MBYTES for IC factorization : 32289</div><div> ** Rank of proc needing largest memory for OOC facto : 0</div>
<div> ** Estimated corresponding MBYTES for OOC facto : 3462</div><div> ** Estimated avg. MBYTES per work. proc at facto (OOC) : 1787</div><div> ** TOTAL space in MBYTES for OOC factorization : 28599</div>
<div>Entering ZMUMPS driver with JOB, N, NZ = 2 122370 5211070</div><div><br></div><div> ****** FACTORIZATION STEP ********</div><div><br></div><div><br></div><div> GLOBAL STATISTICS PRIOR NUMERICAL FACTORIZATION ...</div>
<div> NUMBER OF WORKING PROCESSES = 16</div><div> OUT-OF-CORE OPTION (ICNTL(22)) = 0</div><div> REAL SPACE FOR FACTORS = 565185337</div><div> INTEGER SPACE FOR FACTORS = 3537003</div>
<div> MAXIMUM FRONTAL SIZE (ESTIMATED) = 15239</div><div> NUMBER OF NODES IN THE TREE = 7914</div><div> Convergence error after scaling for ONE-NORM (option 7/8) = 0.79D+00</div><div> Maximum effective relaxed size of S = 199523439</div>
<div> Average effective relaxed size of S = 98303057</div><div><br></div><div> REDISTRIB: TOTAL DATA LOCAL/SENT = 657185 14022665</div><div> GLOBAL TIME FOR MATRIX DISTRIBUTION = 0.4805</div>
<div> ** Memory relaxation parameter ( ICNTL(14) ) : 35</div><div> ** Rank of processor needing largest memory in facto : 0</div><div> ** Space in MBYTES used by this processor for facto : 3661</div>
<div> ** Avg. Space in MBYTES per working proc during facto : 2018</div></div><div><br></div></div></div>