[petsc-users] Extracting data from a Petsc matrix

[Barry Smith]

On Aug 12, 2013, at 4:05 PM, Harshad Sahasrabudhe <hsahasra at purdue.edu> wrote:

> Hi Jed,
> 
> I am now working to add library support for LU decomposition using MAGMA. I need your help with the following:
> 
> 1) How do I add the options --download-magma, --with-magma, etc. to the configure script for building with MAGMA?

   Add a new file in config/PETSc/packages  (copy one that is already there and modify for magma).
> 
> 2) I have a fair idea how the PETSc code is structured and how to add source code to the impls/ directory. How does PETSc get to know that there is an additional implementation (in this case MAGMA) in this directory? Is there a config file of some sort?

    Add the new directory name to list of directories in the makefile in that directory and add in MatRegisterAll().

   Barry

> 
> Thanks,
> Harshad
> 
> ----- Original Message -----
> From: "Jed Brown" <jedbrown at mcs.anl.gov>
> To: hsahasra at purdue.edu, petsc-users at mcs.anl.gov
> Sent: Saturday, July 13, 2013 12:43:08 PM
> Subject: Re: [petsc-users] Extracting data from a Petsc matrix
> 
> "hsahasra at purdue.edu" <hsahasra at purdue.edu> writes:
> 
>> Hi,
>> 
>> I am working on solving a system of linear equations with square
>> matrix. I'm first factoring the matrix using LU decomposition.
> 
> I assume you're solving a dense problem because that is all MAGMA does.
> 
>> I want to do the LU decomposition step using MAGMA on GPUs. MAGMA
>> library implements LAPACK functions on a CPU+GPU based system.
>> 
>> So my question is, how do I extract the data from a Petsc Mat so that
>> it can be sent to the dgetrf routine in MAGMA.
> 
> MatDenseGetArray
> 
>> Is there any need for duplicating the data for this step?
> 
> You're on your own for storage of factors.  Alternatively, you could add
> library support so that you could use PCLU and
> '-pc_factor_mat_solver_package magma' (or PCFactorSetMatSolverPackage).
> Doing this is not a priority for us, but we can provide guidance if you
> want to tackle it.