most efficient way to use local CSR matrix?

Mark, Barry, Thanks for the help. For now I'm sticking with the approach of copying the csr matrix and using the csr data structures to do the preallocations. I'll eventually get around to writing the code for assembling directly into the petsc matrix. I have two more questions. 1) On 1

most efficient way to use local CSR matrix?

On Thu, Oct 29, 2009 at 1:46 PM, Chris Kees christopher.e.kees at usace.army.mil wrote: Mark, Barry, Thanks for the help. For now I'm sticking with the approach of copying the csr matrix and using the csr data structures to do the preallocations. I'll eventually get around to writing the

most efficient way to use local CSR matrix?

I pulled the most recent version of petsc-dev and added superlu_dist. The only case that fails is spooles on 1 processor with an mpiaij matrix created with the Create/SetSizes/... paradigm. I would suspect something in spooles wrapper code. code allocating mat:

most efficient way to use local CSR matrix?

Hmm, this should not happen. The matrix should be identical in both cases (note the initial residual is the same in both cases so they may be identical). Here's one more thing you can try. In the same routine create TWO matrices; one each way, then use MatAXPY() to take the

most efficient way to use local CSR matrix?

I think I added the correct test code and got a difference of 0, but spooles is still not returning the correct result, at least not until the 2nd iteration. I'm attaching some of the code and output. Chris code for mat creates ---

most efficient way to use local CSR matrix?

On Oct 29, 2009, at 5:02 PM, Chris Kees wrote: I think I added the correct test code and got a difference of 0, but spooles is still not returning the correct result, at least not until the 2nd iteration. I'm attaching some of the code and output. We have not been actively updating

most efficient way to use local CSR matrix?

Have you run this with the debug version of PETSc? It does additional argument testing that might catch a funny value passed in. Barry On Oct 29, 2009, at 5:02 PM, Chris Kees wrote: I think I added the correct test code and got a difference of 0, but spooles is still not

most efficient way to use local CSR matrix?

On Oct 29, 2009, at 8:43 PM, Hong Zhang wrote: On Oct 29, 2009, at 5:02 PM, Chris Kees wrote: I think I added the correct test code and got a difference of 0, but spooles is still not returning the correct result, at least not until the 2nd iteration. I'm attaching some of the code

most efficient way to use local CSR matrix?

I'm just using a simple configuration: ./config/configure.py --with-clanguage=C --with-cc='/usr/bin/mpicc - arch x86_64' --with-cxx='/usr/bin/mpicxx -arch x86_64' --without- fortran --with-mpi-compilers --without-shared --without-dynamic -- download-parmeti s=ifneeded

most efficient way to use local CSR matrix?

On Oct 20, 2009, at 10:13 PM, Chris Kees wrote: Thanks. Just to make sure I'm following, when I create the matrix I do: MatCreate(PETSC_COMM_WORLD,self-m); MatSetSizes(self-m,n,n,N,N); MatSetType(self-m,MATMPIAIJ); MatSetFromOptions(self-m);

most efficient way to use local CSR matrix?

Chris, just a note, Perhaps I missed something here but do something similar to you, eg overlapping partitions, and PETSc is setup very well to be intrusive in your code (I sometimes write a little 'addvalue' wrapper) and assemble your FE matrices directly into a global matrix. You use the

most efficient way to use local CSR matrix?

Hi, Our unstructured finite element code does a fairly standard overlapping decomposition that allows it to calculate all the non-zero column entries for the rows owned by the processor (rows 0...ldim-1 in the local numbering). We assemble them into a local CSR matrix and then copy them

most efficient way to use local CSR matrix?

Here's the deal. In parallel PETSc does not use a single CSR matrix on each process to hold the MPIAIJ matrix. Hence if you store the matrix on a process as a single CSR matrix then it has to be copied into the PETSc datastructure. MatMPIAIJSetPreallocationCSR() does the copy very