<div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr">On Sun, Jun 24, 2018 at 2:09 PM Smith, Barry F. <<a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
<br>
> On Jun 24, 2018, at 11:15 AM, Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> wrote:<br>
> <br>
> First, -pc_gamg_threshold 0.8  is crazy. 0.08 is very high. probably best to start with 0.<br>
> <br>
> This threshold will delete all of your edges in the coarsening graph, probably, and GAMG will switch to a one level solve (silently although -info will tell you).<br>
> <br>
> And yes, this looks like a 64 bit indices problem, we have:<br>
> <br>
> #define GAMG_HASH(key) ((((PetscInt)7)*key)%a_tab->size)<br>
> <br>
> We should probably cast everything to (long long) or (unsigned PetscInt) I guess ... I'm not sure how this should be done exactly.  Any ideas from the peanut gallery?<br>
<br>
Mark,<br>
<br>
     It should only be changed if we know the exact mechanism that causes the failure. It is not obvious to me that just changing ((((PetscInt)7)*key)%a_tab->size) to ((((PetscInt64)7)*key)%a_tab->size) should matter.<br></blockquote><div><br></div><div>If 7*key > 2B then you can get a negative number and <span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">%a_tab-></span><span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">size will still be a negative number, which triggers:</span></div><div><span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><br></span></div><div><span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><span style="text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">ksp/pc/impls/gamg/agg.c:          if (flid < 0) SETERRQ(PETSC_COMM_SELF,PETSC_</span><span style="text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">ERR_PLIB,"Cannot find gid1 in table”);</span><br></span></div><div> </div><div>which is the second error that Pierre gets. I don't fully understand the first error but a messed up hash table is certainly a possibility.</div><div><br></div><div>And maybe we should cast the hell out of it to be safe:</div><div><br></div><div>#define GAMG_HASH(key) (<span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">PetscInt</span>)((((PetscInt64)7)*<span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">(PetscInt64)</span>key)%<span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">(PetscInt64)</span>a_tab->size)<br></div><div><br></div><div>I am a bit puzzled that this works. I would think this would be more of a problem but maybe everyone just uses 64 bit ints. I have dealt with this many times and I am puzzled that I left it in a state that looks so wrong to me now. Perhaps Pierre's complicated thing is causing this somehow.</div><div><br></div><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
   Barry<br>
<br>
> <br>
> <br>
> <br>
> On Sun, Jun 24, 2018 at 2:34 AM Pierre Jolivet <<a href="mailto:pierre.jolivet@enseeiht.fr" target="_blank">pierre.jolivet@enseeiht.fr</a>> wrote:<br>
> Hello,<br>
> Any ideas on how to debug this?<br>
> [406]PETSC ERROR: key 328214936 not found in table<br>
> [406]PETSC ERROR: See <a href="http://www.mcs.anl.gov/petsc/documentation/faq.html" rel="noreferrer" target="_blank">http://www.mcs.anl.gov/petsc/documentation/faq.html</a> for trouble shooting.<br>
> [406]PETSC ERROR: Petsc Development GIT revision: v3.9.2-603-gceafe64  GIT Date: 2018-06-10 12:46:16 -0500<br>
> [406]PETSC ERROR: /ccc/work/cont003/rndm/rndm/FreeFem-sources/src/mpi/FreeFem++-mpi on a arch-linux2-c-debug-bullxmpi named curie2698 by jolivetp Sun Jun 24 07:06:07 2018<br>
> [406]PETSC ERROR: Configure options --with-scalar-type=real --with-hypre-dir=arch-linux2-c-opt-bullxmpi --with-metis-dir=arch-linux2-c-opt-bullxmpi --with-mumps-dir=arch-linux2-c-opt-bullxmpi --with-mkl_cpardiso=1 --with-mkl_pardiso=1 --with-parmetis-dir=arch-linux2-c-opt-bullxmpi --with-ptscotch-dir=arch-linux2-c-opt-bullxmpi --with-sowing-dir=arch-linux2-c-opt-bullxmpi --with-blacs-include=/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl/include" rel="noreferrer" target="_blank">18.0.1.163/mkl/include</a> --with-blacs-lib=/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl/lib/intel64/libmkl_blacs_openmpi_lp64.so" rel="noreferrer" target="_blank">18.0.1.163/mkl/lib/intel64/libmkl_blacs_openmpi_lp64.so</a> --with-blaslapack-dir=/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl/lib/intel64" rel="noreferrer" target="_blank">18.0.1.163/mkl/lib/intel64</a> --with-x=0 --with-debugging=1 --with-errorchecking=1 --with-mkl_pardiso-dir=/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl" rel="noreferrer" target="_blank">18.0.1.163/mkl</a> --with-mkl_cpardiso-dir=/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl" rel="noreferrer" target="_blank">18.0.1.163/mkl</a> --with-scalapack-include=/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl/include" rel="noreferrer" target="_blank">18.0.1.163/mkl/include</a> --with-scalapack-lib="[/ccc/products/mkl-18.0.1.163/default/<a href="http://18.0.1.163/mkl/lib/intel64/libmkl_scalapack_lp64.so,/ccc/products/mkl-18.0.1.163/default/18.0.1.163/mkl/lib/intel64/libmkl_blacs_openmpi_lp64.so" rel="noreferrer" target="_blank">18.0.1.163/mkl/lib/intel64/libmkl_scalapack_lp64.so,/ccc/products/mkl-18.0.1.163/default/18.0.1.163/mkl/lib/intel64/libmkl_blacs_openmpi_lp64.so</a>]" --with-fortran-bindings=0 PETSC_ARCH=arch-linux2-c-debug-bullxmpi<br>
> [406]PETSC ERROR: #1 PCGAMGHashTableFind() line 106 in /ccc/work/cont003/rndm/rndm/petsc/include/petsc/private/pcgamgimpl.h<br>
> [406]PETSC ERROR: #2 smoothAggs() line 526 in /ccc/work/cont003/rndm/rndm/petsc/src/ksp/pc/impls/gamg/agg.c<br>
> [406]PETSC ERROR: #3 PCGAMGCoarsen_AGG() line 970 in /ccc/work/cont003/rndm/rndm/petsc/src/ksp/pc/impls/gamg/agg.c<br>
> [406]PETSC ERROR: #4 PCSetUp_GAMG() line 518 in /ccc/work/cont003/rndm/rndm/petsc/src/ksp/pc/impls/gamg/gamg.c<br>
> [406]PETSC ERROR: #5 PCSetUp() line 932 in /ccc/work/cont003/rndm/rndm/petsc/src/ksp/pc/interface/precon.c<br>
> [406]PETSC ERROR: #6 KSPSetUp() line 381 in /ccc/work/cont003/rndm/rndm/petsc/src/ksp/ksp/interface/itfunc.c<br>
> <br>
> For smaller problems, this works fine. Should I compile using 64 bit indices?<br>
> I’m using -pc_type gamg -pc_gamg_threshold 0.8 -pc_gamg_agg_nsmooths 1 -pc_gamg_sym_graph true -pc_gamg_square_graph 10<br>
> <br>
> Thanks,<br>
> Pierre<br>
<br>
</blockquote></div></div>