<div dir="ltr"><div><div>Thank you.<br></div>Runing with <br>mpirun -np 1 valgrind --track-origins=yes ~/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG -ksp_monitor_true_residual -ksp_view<br><br></div>gives the following information. <br>
<div><div><br>==8222== Conditional jump or move depends on uninitialised value(s)<br>==8222== at 0x216E9A7: SearchPath (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>==8222== by 0x216E1EC: mkl_cfg_file (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>
==8222== by 0x216AC14: DDOT (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>==8222== by 0x1291100: VecNorm_Seq (bvec2.c:239)<br>==8222== by 0x126F431: VecNorm (rvector.c:166)<br>==8222== by 0x127039D: VecNormalize (rvector.c:261)<br>
==8222== by 0x1405507: KSPGMRESCycle (gmres.c:127)<br>==8222== by 0x140695F: KSPSolve_GMRES (gmres.c:231)<br>==8222== by 0x1BEEF66: KSPSolve (itfunc.c:446)<br>==8222== by 0x13F95E8: kspsolve_ (itfuncf.c:219)<br>
==8222== by 0xC5E51F: petsolv_ (PETSOLV.F:375)<br>==8222== by 0x612C35: flowsol_ng_ (flowsol_ng.F:275)<br>==8222== Uninitialised value was created by a stack allocation<br>==8222== at 0x216E97F: SearchPath (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>
==8222== <br>==8222== Conditional jump or move depends on uninitialised value(s)<br>==8222== at 0x216E9D1: SearchPath (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>==8222== by 0x216E1EC: mkl_cfg_file (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>
==8222== by 0x216AC14: DDOT (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>==8222== by 0x1291100: VecNorm_Seq (bvec2.c:239)<br>==8222== by 0x126F431: VecNorm (rvector.c:166)<br>==8222== by 0x127039D: VecNormalize (rvector.c:261)<br>
==8222== by 0x1405507: KSPGMRESCycle (gmres.c:127)<br>==8222== by 0x140695F: KSPSolve_GMRES (gmres.c:231)<br>==8222== by 0x1BEEF66: KSPSolve (itfunc.c:446)<br>==8222== by 0x13F95E8: kspsolve_ (itfuncf.c:219)<br>
==8222== by 0xC5E51F: petsolv_ (PETSOLV.F:375)<br>==8222== by 0x612C35: flowsol_ng_ (flowsol_ng.F:275)<br>==8222== Uninitialised value was created by a stack allocation<br>==8222== at 0x216E97F: SearchPath (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64_DEBUG)<br>
==8222== <br><br></div></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, May 5, 2014 at 10:03 AM, Barry Smith <span dir="ltr"><<a href="mailto:bsmith@mcs.anl.gov" target="_blank">bsmith@mcs.anl.gov</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
If you run valgrind with the debug version of the libraries it will provide more information about the line numbers where the problem occurred, etc. recommend doing that.<br>
<br>
Either your initial solution or right hand side has garbage in it or the wrong blas may be being linked in. But there is definitely a problem<br>
<span class="HOEnZb"><font color="#888888"><br>
Barry<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
On May 5, 2014, at 8:28 AM, Song Gao <<a href="mailto:song.gao2@mail.mcgill.ca">song.gao2@mail.mcgill.ca</a>> wrote:<br>
<br>
> Thanks for reply. What do you mean by a “happier” state? I check the converged solution (the one which call kspgmressetrestart twice), the solution should be correct.<br>
><br>
> I run with valgrind both codes (one call kspgmressetrestart once and another call kspgmressetrestart twice)<br>
> Both of them have the errors: what does this mean? Thank you in advance.<br>
> ==7858== Conditional jump or move depends on uninitialised value(s)<br>
> ==7858== at 0xE71DFB: SearchPath (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0xE71640: mkl_cfg_file (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0xE6E068: DDOT (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x73281A: VecNorm_Seq (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x730BF4: VecNormalize (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x7BC5A8: KSPSolve_GMRES (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0xB8A06E: KSPSolve (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x7B659F: kspsolve_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x5EAE84: petsolv_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x4ECD46: flowsol_ng_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x507E4E: iterprc_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x51D1B4: solnalg_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858==<br>
> ==7858== Conditional jump or move depends on uninitialised value(s)<br>
> ==7858== at 0xE71E25: SearchPath (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0xE71640: mkl_cfg_file (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0xE6E068: DDOT (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x73281A: VecNorm_Seq (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x730BF4: VecNormalize (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x7BC5A8: KSPSolve_GMRES (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0xB8A06E: KSPSolve (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x7B659F: kspsolve_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x5EAE84: petsolv_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x4ECD46: flowsol_ng_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x507E4E: iterprc_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858== by 0x51D1B4: solnalg_ (in /home/cfd/sgao/mycodes/fensapngnew/bin/fensapMPI_LINUX64)<br>
> ==7858==<br>
><br>
><br>
> On Fri, May 2, 2014 at 7:25 PM, Barry Smith <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>> wrote:<br>
><br>
> On May 2, 2014, at 5:29 PM, Song Gao <<a href="mailto:song.gao2@mail.mcgill.ca">song.gao2@mail.mcgill.ca</a>> wrote:<br>
><br>
> > Thanks for your quick reply. What confused me is that why would the code works fine if I reset the gmres restart number by recalling kspgmressetrestart just before kspsolve?<br>
><br>
> It isn’t really working. Something is going wrong (run with valgrind) and setting that restart number and starting the solver just puts it in a “happier” state so it seems to make more progress.<br>
><br>
> Barry<br>
><br>
> ><br>
> > Sent from my iPhone<br>
> ><br>
> >> On May 2, 2014, at 6:03 PM, "Barry Smith" <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>> wrote:<br>
> >><br>
> >><br>
> >> Your shell matrix is buggy in some way. Whenever the residual norm jumps like crazy at a restart it means that something is wrong with the operator.<br>
> >><br>
> >> Barry<br>
> >><br>
> >>> On May 2, 2014, at 4:41 PM, Song Gao <<a href="mailto:song.gao2@mail.mcgill.ca">song.gao2@mail.mcgill.ca</a>> wrote:<br>
> >>><br>
> >>> Dear PETSc users,<br>
> >>><br>
> >>> I'm solving a linear system in KSP and trying to setup the solver in codes. But I feel strange because my codes don't converge unless I call KSPGMRESSetRestart twice.<br>
> >>><br>
> >>> My codes looks like<br>
> >>><br>
> >>> call KSPSetOperators ( pet_solv, pet_mat_mf_shell, pet_matp, DIFFERENT_NONZERO_PATTERN, ierpetsc )<br>
> >>> call KSPSetType ( pet_solv, 'gmres', ierpetsc )<br>
> >>> call KSPGMRESSetRestart ( pet_solv, 30, ierpetsc )<br>
> >>> call KSPGetPC ( pet_solv, pet_precon, ierpetsc )<br>
> >>> call PCSetType ( pet_precon, 'asm', ierpetsc )<br>
> >>> call PCASMSetOverlap ( pet_precon, 1, ierpetsc )<br>
> >>> call KSPSetUp ( pet_solv, ierpetsc )<br>
> >>> call PCASMGetSubKSP ( pet_precon, n_local, first_local, pet_solv_sub, ierpetsc ) ! n_local is one<br>
> >>> call KSPGetPC ( pet_solv_sub(1), pet_precon_sub, ierpetsc )<br>
> >>> call PCSetType ( pet_precon_sub, 'jacobi', ierpetsc )<br>
> >>> call PCJacobiSetUseRowMax ( pet_precon_sub, ierpetsc )<br>
> >>> call KSPSetFromOptions ( pet_solv, ierpetsc )<br>
> >>> call KSPGMRESSetRestart ( pet_solv, 29, ierpetsc ) ! adding this line, the codes converge<br>
> >>> call KSPSolve ( pet_solv, pet_rhsp, pet_solup, ierpetsc )<br>
> >>><br>
> >>> runing with 1 CPU WITHOUT the line with red color and the codes don't converge<br>
> >>><br>
> >>> runtime options: -ksp_monitor_true_residual -ksp_view<br>
> >>> 0 KSP preconditioned resid norm 6.585278940829e+00 true resid norm 9.619278462343e-03 ||r(i)||/||b|| 1.000000000000e+00<br>
> >>> 1 KSP preconditioned resid norm 6.585278219510e+00 true resid norm 9.619278462343e-03 ||r(i)||/||b|| 1.000000000000e+00<br>
> >>> 2 KSP preconditioned resid norm 2.198638170622e+00 true resid norm 1.365132713014e-01 ||r(i)||/||b|| 1.419163317039e+01<br>
> >>> 3 KSP preconditioned resid norm 1.599896387215e+00 true resid norm 1.445988845022e-01 ||r(i)||/||b|| 1.503219654865e+01<br>
> >>> .......<br>
> >>> 28 KSP preconditioned resid norm 4.478466011191e-01 true resid norm 1.529879309381e-01 ||r(i)||/||b|| 1.590430420920e+01<br>
> >>> 29 KSP preconditioned resid norm 4.398129572260e-01 true resid norm 1.530132924055e-01 ||r(i)||/||b|| 1.590694073413e+01<br>
> >>> 30 KSP preconditioned resid norm 2.783227613716e+12 true resid norm 1.530369123550e-01 ||r(i)||/||b|| 1.590939621450e+01<br>
> >>><br>
> >>> KSP Object: 1 MPI processes<br>
> >>> type: gmres<br>
> >>> GMRES: restart=30, using Classical (unmodified) Gram-Schmidt Orthogonalization with no iterative refinement<br>
> >>> GMRES: happy breakdown tolerance 1e-30<br>
> >>> maximum iterations=10000, initial guess is zero<br>
> >>> tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
> >>> left preconditioning<br>
> >>> using PRECONDITIONED norm type for convergence test<br>
> >>> PC Object: 1 MPI processes<br>
> >>> type: asm<br>
> >>> Additive Schwarz: total subdomain blocks = 1, amount of overlap = 1<br>
> >>> Additive Schwarz: restriction/interpolation type - RESTRICT<br>
> >>> [0] number of local blocks = 1<br>
> >>> Local solve info for each block is in the following KSP and PC objects:<br>
> >>> - - - - - - - - - - - - - - - - - -<br>
> >>> [0] local block number 0, size = 22905<br>
> >>> KSP Object: (sub_) 1 MPI processes<br>
> >>> type: preonly<br>
> >>> maximum iterations=10000, initial guess is zero<br>
> >>> tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
> >>> left preconditioning<br>
> >>> using NONE norm type for convergence test<br>
> >>> PC Object: (sub_) 1 MPI processes<br>
> >>> type: jacobi<br>
> >>> linear system matrix = precond matrix:<br>
> >>> Matrix Object: 1 MPI processes<br>
> >>> type: seqbaij<br>
> >>> rows=22905, cols=22905, bs=5<br>
> >>> total: nonzeros=785525, allocated nonzeros=785525<br>
> >>> total number of mallocs used during MatSetValues calls =0<br>
> >>> block size is 5<br>
> >>> - - - - - - - - - - - - - - - - - -<br>
> >>> linear system matrix followed by preconditioner matrix:<br>
> >>> Matrix Object: 1 MPI processes<br>
> >>> type: shell<br>
> >>> rows=22905, cols=22905<br>
> >>> Matrix Object: 1 MPI processes<br>
> >>> type: seqbaij<br>
> >>> rows=22905, cols=22905, bs=5<br>
> >>> total: nonzeros=785525, allocated nonzeros=785525<br>
> >>> total number of mallocs used during MatSetValues calls =0<br>
> >>> block size is 5<br>
> >>> WARNING: zero iteration in iterative solver<br>
> >>><br>
> >>> runing with 1 CPU WITH the line with red color and the codes converge<br>
> >>><br>
> >>> runtime options: -ksp_monitor_true_residual -ksp_view<br>
> >>> 0 KSP preconditioned resid norm 6.585278940829e+00 true resid norm 9.619278462343e-03 ||r(i)||/||b|| 1.000000000000e+00<br>
> >>> 1 KSP preconditioned resid norm 2.566248171026e+00 true resid norm 4.841043870812e-03 ||r(i)||/||b|| 5.032647604250e-01<br>
> >>> 2 KSP preconditioned resid norm 1.410418402651e+00 true resid norm 3.347509391208e-03 ||r(i)||/||b|| 3.480000505561e-01<br>
> >>> 3 KSP preconditioned resid norm 9.665409287757e-01 true resid norm 2.289877121679e-03 ||r(i)||/||b|| 2.380508195748e-01<br>
> >>> 4 KSP preconditioned resid norm 4.469486152454e-01 true resid norm 1.283813398084e-03 ||r(i)||/||b|| 1.334625463968e-01<br>
> >>> 5 KSP preconditioned resid norm 2.474889829653e-01 true resid norm 7.956009139680e-04 ||r(i)||/||b|| 8.270900120862e-02<br>
> >>> ............<br>
> >>> 24 KSP preconditioned resid norm 9.518780877620e-05 true resid norm 6.273993696172e-07 ||r(i)||/||b|| 6.522312167937e-05<br>
> >>> 25 KSP preconditioned resid norm 6.837876679998e-05 true resid norm 4.612861071815e-07 ||r(i)||/||b|| 4.795433555514e-05<br>
> >>> 26 KSP preconditioned resid norm 4.864361942316e-05 true resid norm 3.394754589076e-07 ||r(i)||/||b|| 3.529115621682e-05<br>
> >>> KSP Object: 1 MPI processes<br>
> >>> type: gmres<br>
> >>> GMRES: restart=29, using Classical (unmodified) Gram-Schmidt Orthogonalization with no iterative refinement<br>
> >>> GMRES: happy breakdown tolerance 1e-30<br>
> >>> maximum iterations=10000, initial guess is zero<br>
> >>> tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
> >>> left preconditioning<br>
> >>> using PRECONDITIONED norm type for convergence test<br>
> >>> PC Object: 1 MPI processes<br>
> >>> type: asm<br>
> >>> Additive Schwarz: total subdomain blocks = 1, amount of overlap = 1<br>
> >>> Additive Schwarz: restriction/interpolation type - RESTRICT<br>
> >>> [0] number of local blocks = 1<br>
> >>> Local solve info for each block is in the following KSP and PC objects:<br>
> >>> - - - - - - - - - - - - - - - - - -<br>
> >>> [0] local block number 0, size = 22905<br>
> >>> KSP Object: (sub_) 1 MPI processes<br>
> >>> type: preonly<br>
> >>> maximum iterations=10000, initial guess is zero<br>
> >>> tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
> >>> left preconditioning<br>
> >>> using NONE norm type for convergence test<br>
> >>> PC Object: (sub_) 1 MPI processes<br>
> >>> type: jacobi<br>
> >>> linear system matrix = precond matrix:<br>
> >>> Matrix Object: 1 MPI processes<br>
> >>> type: seqbaij<br>
> >>> rows=22905, cols=22905, bs=5<br>
> >>> total: nonzeros=785525, allocated nonzeros=785525<br>
> >>> total number of mallocs used during MatSetValues calls =0<br>
> >>> block size is 5<br>
> >>> - - - - - - - - - - - - - - - - - -<br>
> >>> linear system matrix followed by preconditioner matrix:<br>
> >>> Matrix Object: 1 MPI processes<br>
> >>> type: shell<br>
> >>> rows=22905, cols=22905<br>
> >>> Matrix Object: 1 MPI processes<br>
> >>> type: seqbaij<br>
> >>> rows=22905, cols=22905, bs=5<br>
> >>> total: nonzeros=785525, allocated nonzeros=785525<br>
> >>> total number of mallocs used during MatSetValues calls =0<br>
> >>> block size is 5<br>
> >>> WARNING: zero iteration in iterative solver<br>
> >>><br>
> >>><br>
> >>> What would be my error here? Thank you.<br>
> >><br>
><br>
><br>
<br>
</div></div></blockquote></div><br></div>