November 2010 - petsc-users - mailman3.cels.anl.gov

Re: [petsc-users] Can I use MAT_FINAL_ASSEMBLY instead of MAT_FLUSH_ASSEMBLY?
by Rongliang Chen 09 Nov '10

09 Nov '10

Message: 2 > Date: Mon, 8 Nov 2010 19:57:16 -0600 > From: Barry Smith <bsmith(a)mcs.anl.gov> > Subject: Re: [petsc-users] Can I use MAT_FINAL_ASSEMBLY instead of > MAT_FLUSH_ASSEMBLY? > To: PETSc users list <petsc-users(a)mcs.anl.gov> > Message-ID: <52AFD8FB-17C1-4B51-97E9-C77112B69F3C(a)mcs.anl.gov> > Content-Type: text/plain; charset=us-ascii > > > On Nov 8, 2010, at 7:49 PM, Rongliang Chen wrote: > > > Hi everyone, > > > > Can I use MAT_FINAL_ASSEMBLY instead of MAT_FLUSH_ASSEMBLY in the > function MatAssemblyBegin/End when switching between ADD_VALUES and > INSERT_VALUES in MatSetValues()? Thank you! > > > > Regards, > > > > Rongliang > > Yes, but generally there is no reason to. It is best to use the flush until > your final assembly because then it doesn't do any work that needs to be > redone or prematurely free memory that may be needed later. > > Why do you want to do that? > > Barry > > > Hi Barry, Thank you for your reply. I find that when I use MAT_FINAL_ASSEMBLY instead of MAT_FLUSH_ASSEMBLY in the function MatAssemblyBegin/End, it will cause problem. I do not know why. The reason that I want to use it like this is that I first set the matrix with ADD_VALUES and then I need to zero some rows of the matrix using MatZeroRowsLocalIS(), at last I still need to set some elements of the matrix with INSERT_VALUES. Between MatSetValues() and MatZeroRowsLocalIS() I need to use MAT_FINAL_ASSEMBLY. The order is: ADD_VALUES--->MatZeroRowsLocalIS()--->INSERT_VALUES. Thanks! Regards, Rongliang

2 3

Can I use MAT_FINAL_ASSEMBLY instead of MAT_FLUSH_ASSEMBLY?
by Rongliang Chen 08 Nov '10

08 Nov '10

Hi everyone, Can I use MAT_FINAL_ASSEMBLY instead of MAT_FLUSH_ASSEMBLY in the function MatAssemblyBegin/End when switching between ADD_VALUES and INSERT_VALUES in MatSetValues()? Thank you! Regards, Rongliang

2 1

MatCreateMPIAIJWithSplitArrays for symmetric matrix?
by Daniel Langr 08 Nov '10

08 Nov '10

Hi all, Is there an equivalent of MatCreateMPIAIJWithSplitArrays function for symmetric matrices? Or does this function work with symmetric matrices? I work with symmetric matrices that are extremely expensive to compute and I don't know their structure in advance. So to do preallocation, it is not practical to compute a matrix twice - firstly only to get d_nnz and o_nnz arrays and secondly to set values. Now, I solve this problem by storing a matrix to a file system, which is faster but still far from ideal. The best solution I see is to compute a matrix only once, store it in memory as split CSR arrays and then just call something like MatCreateMPISBAIJWithSplitArrays. Thanks, Daniel

2 10

Mixed 32-bit and 64-bit indices
by Daniel Langr 08 Nov '10

08 Nov '10

Hi all, is there any way to use 32-bit integers for indexing rows/columns and 64-bit integers for indexing nonzeros at the same time in PETSc? I have a matrix with 4,559,168,066 nonzeros but only 448,797 rows/columns. For a general CSR scheme, the memory savings are almost 25% when using 32-bit integers compared to 64-bit for indexing columns, which is really a lot of memory. Thanks, Daniel

2 2

No warning or error message of segmentation violation
by Li, Zhisong (lizs) 05 Nov '10

05 Nov '10

Hi, Petsc Team, I use DA object to create structured data array based on global vectors. Recently, I checked my code and found an array with index out of range. But the code can run without any warning or segmentation violation message. At first I thought this might be a cache problem. But later I ran a simpler test code on another machine and it gave no error either. I did not use any periodic boundary setting. This never took place in the past from my experience, as Petsc could always detect any segmentation violation. So what's wrong with it this time? The following is the main part of my sequential test code: DACreate3d(PETSC_COMM_WORLD,DA_NONPERIODIC,DA_STENCIL_BOX,m,n,q,PETSC_DECIDE,PETSC_DECIDE,PETSC_DECIDE,1,1,PETSC_NULL,PETSC_NULL,PETSC_NULL,&da); DACreateGlobalVector(da, &elev); VecSet(elev, 3.14159265); DAVecGetArray(da, elev, &ele); PetscPrintf(PETSC_COMM_WORLD, "ele[3][3][%D] is %4.3e\n", m+120, ele[3][3][m+120]); DAVecRestoreArray(da, elev, &ele); VecDestroy(elev); DADestroy(daw); Thank you very much. Zhisong Li

3 2

Re: [petsc-users] Initial guess for the GMRES solver in PETSC
by jinshan wu 05 Nov '10

05 Nov '10

Aron and Dave, thanks both of you a lot. Jinshan Message: 6 Date: Fri, 5 Nov 2010 08:54:15 +0100 From: Dave May <dave.mayhem23(a)gmail.com> Subject: Re: [petsc-users] Initial guess for the GMRES solver in PETSC To: PETSc users list <petsc-users(a)mcs.anl.gov> Message-ID: <AANLkTikzKoBnD9n+dkMrUmR3FUry_BhgEGFXO8cZJfk-(a)mail.gmail.com> Content-Type: text/plain; charset=ISO-8859-1 Or you can specify this directly in your code. See http://www.mcs.anl.gov/petsc/petsc-as/snapshots/petsc-current/docs/manualpa… Cheers, Dave On 5 November 2010 07:11, Aron Ahmadia <aron.ahmadia(a)kaust.edu.sa> wrote: > Hi Jinshan, > > You can pass your initial guess in to the KSP object as X, then at the > command line set the flag: > > -ksp_initial_guess_nonzero > > Happy Computing, > Aron > > On Fri, Nov 5, 2010 at 8:41 AM, jinshan wu <wujinshan(a)yahoo.com> wrote: >> Hi all, >> >> I am using the GMRES linear solver in PETSC. I have a very good initial guess >>of >> the true solution. Can I somehow supply it to the solver? It should help the >> solver speed up the convergence if I can do that. Thanks. >> >> Jinshan >> >> >> >> > ------------------------------ _______________________________________________ petsc-users mailing list petsc-users(a)mcs.anl.gov https://lists.mcs.anl.gov/mailman/listinfo/petsc-users End of petsc-users Digest, Vol 23, Issue 4 ******************************************

1 0

Initial guess for the GMRES solver in PETSC
by jinshan wu 05 Nov '10

05 Nov '10

Hi all, I am using the GMRES linear solver in PETSC. I have a very good initial guess of the true solution. Can I somehow supply it to the solver? It should help the solver speed up the convergence if I can do that. Thanks. Jinshan

3 2

Petsc with Intel Fortran
by Sharat N 05 Nov '10

05 Nov '10

Hi, I configured the petsc system with intel fortran compiler. This time when I did 'make' on any fortran example. It compiled but gave me 1 warning and 1 error. 1) *warning: feupdateenv is not implemented and will always fail* I found out that This warning is due to a mismatch of library types and can be ignored. The warning will not appear if -shared-intel is used. 2) *symbol lookup error: /opt/intel/Compiler/11.1/072/lib/intel64/libifcore.so.5: undefined symbol: __intel_sse2_strlen* * * *This error is not executing my program does anyone know what is the reasons for this error and how can we solve it?* -- Sharat

2 1

Re: [petsc-users] Boomeramg as a solver
by jibum kim 04 Nov '10

04 Nov '10

Now it is clear. Thank you for your help. - Jibum ----- Original Message ----- From: "jibum kim" <jzk164(a)psu.edu> To: "PETSc users list" <petsc-users(a)mcs.anl.gov> Sent: Monday, November 01, 2010 3:46 PM Subject: Re: [petsc-users] Boomeramg as a solver > Are you saying that boomeramg with richardson iterations is same as the > multigrid without a preconditioner? and if it the case, how can I set an > extra preconditioner like ILU ? > > Thanks, > > - Jibum > > ----- Original Message ----- > From: "Barry Smith" <bsmith(a)mcs.anl.gov> > To: "PETSc users list" <petsc-users(a)mcs.anl.gov> > Sent: Monday, November 01, 2010 2:25 PM > Subject: Re: [petsc-users] Boomeramg as a solver > > > > On Nov 1, 2010, at 12:35 PM, jibum kim wrote: > >> Hello, >> >> I've used Boomeramg as a preconditioner. >> >> Now, I am wondering whether I can use Boomeramg as a solver ? and is it >> possible that I use Boomeramg as both a preconditioner and a solver? >> >> I saw one slide saying that I can use KSPSetType(ksp, KSPRICHARDSON); >> >> But when I try it and use -ksp_view, it shows that my solver is >> Richardson not Boomeramg. >> >> Could u clarify this issue ? > > Richardson is our name for the iterative process u^{n+1} = u^n + B(b - > A u^n) for n=1,...max_it where B is the application of the preconditioner > (in other words running a linear iterative method without any Krylov > acceleration). In BoomerAMG's case B is one application of the V or W > cycle (depending on your choice of BoomerAMG options). So even though it > says the "solver" is Richardson it is using BoomerAMG for the entire > solution process, that is algebraic multigrid with multiple V (or W > cycles). > > Barry > > > > >> >> Thank you. >> >> - Jibum > >

6 16

Boomeramg as a solver
by jibum kim 01 Nov '10

01 Nov '10

Hello, I've used Boomeramg as a preconditioner. Now, I am wondering whether I can use Boomeramg as a solver ? and is it possible that I use Boomeramg as both a preconditioner and a solver? I saw one slide saying that I can use KSPSetType(ksp, KSPRICHARDSON); But when I try it and use -ksp_view, it shows that my solver is Richardson not Boomeramg. Could u clarify this issue ? Thank you. - Jibum

3 4