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

Adding adaptive mesh refinement to current fortran code
by TAY wee-beng 18 Nov '10

18 Nov '10

Hi, I am thinking of adding adaptive mesh refinement to my current fortran code. I am using an immersed boundary code with staggered cartesian grid. I am using PETSc to solve my momentum equations and PETSc/hypre to solve my poisson equation. The code now runs in parallel. Since I'm programming in fortran, some other c++ packages may not be feasible. I am considering packages like paramesh and AGRIF. I wonder if it is possible to integrate them together with PETSc. Is there anything which I need to consider too? Thanks! -- Yours sincerely, TAY wee-beng

2 3

64-bit integer options
by Daniel Langr 17 Nov '10

17 Nov '10

Hi all, I have a problem passing an integer option value greater than 2^32 to my program. I compiled PETSc with --with-64-bit-indices=1, but it seems PetscOptionsGetInt() uses aoti() internally; I would expect atol() or another function named PetscOptionsGetLong(). Thanks, Daniel

4 10

Re: [petsc-users] dof>1 vector tools
by Edward Bueler 16 Nov '10

16 Nov '10

Jed-- > ... > > I guess I would want the loop to be replaced by > > "DAVecScaleDOF(da,v,k,alpha)" for k=0,...,dof-1. > > It's spelled VecStrideScale(). > > > Same question for VecNorm: norm just an identified component? > > You guessed it, VecStrideNorm() RIght! Sorry to miss that. Didn't search on "Stride". > But, I actually recommend nondimensionalizing ... so that these component > norms are not needed by the solvers, as the nondimensional system will > generally produce a better conditioned algebraic system. Did not need the norm for the solvers, actually. I wrote a scaling/descaling "shell" around the residual evaluation routine, so the latter could be dimensional. At this early, prototype stage, where I am at on my code, I don't want the "clutter" of nondimensionalizing PDEs if I am familiar with them in dimensional units. > ... nondimensionalizing (in the sense of ex48, not in terms on > nondimensional numbers ...) ... ex48.c is heavy-going if what one wants is an example showing how PETSc can help with scaling the residual and Jacobian for better conditioning. Related question: Is it true that an instance of DALocalFunction1 will cause trouble if it modifies the solution array? (The values supplied in the "Field *u" argument?) It seems that this is so, which I did not expect for this call-back. Related comment: There is no man page on DALocalFunction1, I think. Thanks! Ed -- ********************************** Ed Bueler 474-7693 (fax 474-5394) Dept of Mathematics and Statistics 301C Chapman Hall, Univ of Alaska Fairbanks, AK 99775-6660 www.dms.uaf.edu/~bueler **********************************

2 1

Re: [petsc-users] sending configure.log to mailing lists
by Satish Balay 15 Nov '10

15 Nov '10

On Mon, 15 Nov 2010, Srinath Vadlamani wrote: > Thanks for checking. The team does want configure.log with all help > requests, correct? yeah - all issues requiring configure.log should be sent to petsc-maint. There is some confusion regarding public mailing lists [petsc-dev, petsc-users] and private list [petsc-maint]. Here is some clarification: 1. To reach petsc-developers - any of the above e-mail ids are equivalent, although petsc-dev is supporsed to be for discussions on development version of petsc. 2. We do not want to flood inboxes of all mailing list subscribers with multi-megabyte configure.log or similar attachments. For this reason - any issues that involve configure.log [esp installation issues] are best sent to petsc-maint. 3. Mailing lists [petsc-dev,petsc-users] need subscription - but no such thing for petsc-maint]. Also mailing list emails reach all folks - while petsc-maint reaches only developers. And mailing lists conversation is archived and accessiable by all for future searches. 3. We do want to keep both the public and private conversation channels open - hence the existance of mailing lists as well as petsc-maint. Satish

1 0

I want to quit PETSc users list
by 陈乐平 15 Nov '10

15 Nov '10

I want to quit PETSc users list. Thank you! ________________________________

2 1

dof>1 vector tools
by Edward Bueler 14 Nov '10

14 Nov '10

Dear PETSc-- This is an actual user request. I am working on a typical PETSc task involving two nonlinear PDEs coupled together. I notice that there are several tools for Vecs which don't help much when I have a dof>1 Vec from my DA, for which the components have different scaling. Thus VecNorm and VecScale are not really what I want. Is there a plan for component-wise versions of those? For example, is there a better/quicker way to scale one component of a dof>1 DA-derived Vec than writing my own, something like this: typedef struct { PetscReal H, u; } Node; ierr = DACreate1d(PETSC_COMM_WORLD,DA_NONPERIODIC,-M,2,1,PETSC_NULL,&da);CHKERRQ(ierr); ierr = DACreateGlobalVector(da,&v);CHKERRQ(ierr); ... Node *n; ierr = DAVecGetArray(da,v,&n);CHKERRQ(ierr); for (i = xs; i < xs + xm; i++) { n[i].u *= alpha; } ierr = DAVecRestoreArray(da,v,&n);CHKERRQ(ierr); I guess I would want the loop to be replaced by "DAVecScaleDOF(da,v,k,alpha)" for k=0,...,dof-1. Same question for VecNorm: norm just an identified component? Ed -- ********************************** Ed Bueler 474-7693 (fax 474-5394) Dept of Mathematics and Statistics 301C Chapman Hall, Univ of Alaska Fairbanks, AK 99775-6660 www.dms.uaf.edu/~bueler **********************************

2 1

minres and cholesky or icc preconditioner
by Umut Tabak 14 Nov '10

14 Nov '10

Dear all, I am trying to solve some linear systems with minres where the operator matrices can become indefinite depending on \lambda and moreover both the operator matrices and the right hand side is changing in a loop. A = K - \lambda M; I used the basic example, ex1.c, and tried supply some command line arguments to see what is going on for the linear solves with minres. Use of -pc_type cholesky or icc does not return an error however I could not understand this because the operator matrix is not PD(symmetric however) so I did not get how cholesky or icc might be used as a preconditioner in this case. However, the results are fine, but the iteration numbers are a bit high maybe I should play with the tolerance, because I am not looking for the very close approximate solution, a coarse representation of the solution might also help me. This is because I would like to use these solution vectors in some other context mixed with some other vectors. And the other question is that backslash of MATLAB solves these systems faster than umfpack or mumps interface of PETSc, is this due to the fact that my PETSc configuration is the default debug one(I read somewhere in the documentation some time ago, mentioning that there can be important performance differences, but could not locate that right now.)? I would appreciate clarification on these points. Best, Umut -- - Hope is a good thing, maybe the best of things and no good thing ever dies... The Shawshank Redemption, replique of Tim Robbins

2 2

Logging the same pieces of code several times using PetscLogEventBegin/PetscLogEventEnd
by Filippo Spiga 11 Nov '10

11 Nov '10

Dear all, I would like to timing the same pieces of code several times using PetscLogEventBegin/PetscLogEventEnd. My code does, more or less, something like that PetscLogEvent logA, logB; PetscLogEventRegister( "Log A", 0, &logA); PetscLogEventRegister( "Log B", 0, &logB); for (i=0...10) { PetscLogEventBegin(logA, 0, 0, 0, 0); ... PetscLogEventEnd(logA, 0, 0, 0, 0); PetscLogEventBegin(logB 0, 0, 0, 0); .. PetscLogEventEnd(logB, 0, 0, 0, 0); } At the end, when I print the summary of all the evens using -log_summary, "Log A" and "Log B" appear many times... as many times as I performed the loop. How can change this behaviour? Thank you very much in advance! Cheers -- Filippo SPIGA, MSc Computer Science

2 2

MATAIJ solvers
by Chris Bording 11 Nov '10

11 Nov '10

Hi I am new to petsc and am attempting to parallelize ex13.c (Poisson example) in the ksp tutorials. I have been able to write some simple test codes and get them to work correctly and learn some of the mechanics of what petsc is doing, with respect to creating mpi vectors and matrices and in using the DA functions. I attempting to test the different solvers I am having some trouble... When I try to run with the same solver it fails with a message about Matrix format for mpiaij does not have a built-in PETSc direct solver. Which solvers are available for mpiaij formats? Also I created the Matrix and Vectors using the DACreate2d(dA,PETSC_COMM_WORLD,..... DAGetMatrix(dA,MATMPIAIJ,&A) DACreateGlobalVector(dA,&x) DACreateGlobalVector(dA,&b) .... is this dictating the types of solvers available? Is there another example of the variable Poisson in Parallel? or creating in a variable RHS and the rho variable in ex13 that maps from local to global indices? Thanks Chris Bording

2 1

Index set with zero size
by Tim Kroeger 11 Nov '10

11 Nov '10

Dear all, Is it allowed to work with an index set (IS) with size 0? In particular, using such an index set in VecScatter and MatGetSubMatrix, I am creating the following objects: v : a vector of size n>0, w : a vector of size 0, A : a matrix of size n\times0. Then I do MatMultAdd (A, w, v, v); This actually seems to change the value of v, although I supposed it not to do so. Are such index sets allowed or not? If not, what about index sets having local size 0 on some processors? Best Regards, Tim -- Dr. Tim Kroeger CeVis -- Center of Complex Systems and Visualization University of Bremen tim.kroeger(a)cevis.uni-bremen.de Universitaetsallee 29 tim.kroeger(a)mevis.fraunhofer.de D-28359 Bremen Phone +49-421-218-7710 Germany Fax +49-421-218-4236

2 12