<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Wed, Jun 10, 2015 at 10:31 AM, Tim Steinhoff <span dir="ltr"><<a href="mailto:kandanovian@gmail.com" target="_blank">kandanovian@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">One can switch on-the-fly?<br></blockquote><div><br></div><div>Yes, either by changing the string in MatSetType() or calling MatSetFromOptions() and passing</div><div><br></div><div> -mat_type baij</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Say,<br>
<br>
we have a block sparse matrix A initilized via MatSetType(A,MATBAIJ)<br>
and subsequent calls of MatSeqBAIJSetPreallocation and<br>
MatSetValuesBlocked<br></blockquote><div><br></div><div>The type determines the storage format, so you do not really want to change after construction. YOu</div><div>can do this however using MatConvert().</div><div><br></div><div> Thanks,</div><div><br></div><div> Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
for use with e.g. the built-in Krylov subspace methods and afterwards<br>
changing type via MatSetType(A,MATAIJ)?<br>
<br>
Regards<br>
Tim<br>
<br>
2015-06-10 17:24 GMT+02:00 Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>>:<br>
> On Wed, Jun 10, 2015 at 10:20 AM, Tim Steinhoff <<a href="mailto:kandanovian@gmail.com">kandanovian@gmail.com</a>><br>
> wrote:<br>
>><br>
>> I wanted to give the multi frontal LU factorization approach a shot as<br>
>> an example for a direct sparse solver.<br>
><br>
><br>
> You can just switch types when you use something other than UMFPACK.<br>
><br>
> Thanks,<br>
><br>
> Matt<br>
><br>
>><br>
>> Regards<br>
>> Tim<br>
>><br>
>> 2015-06-10 17:00 GMT+02:00 Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>>:<br>
>> > On Wed, Jun 10, 2015 at 9:55 AM, Tim Steinhoff <<a href="mailto:kandanovian@gmail.com">kandanovian@gmail.com</a>><br>
>> > wrote:<br>
>> >><br>
>> >> Thanks for the quick reply!<br>
>> >><br>
>> >> "Yes, however in order to get improved performance, you need type<br>
>> >> MATBAIJ."<br>
>> >><br>
>> >> I considered MatSetType(A,MATAIJ); i.e. the non-block type since<br>
>> >> UMFPACK seems to require the seqaij type according to the summary<br>
>> >> page. So do I have to refrain from using the more amiable block-type<br>
>> >> if I want to make use of UMFPACK?<br>
>> ><br>
>> ><br>
>> > Yes, I think so. What do you need in UMFPACK?<br>
>> ><br>
>> > Thanks,<br>
>> ><br>
>> > Matt<br>
>> ><br>
>> >><br>
>> >> 2015-06-10 16:45 GMT+02:00 Matthew Knepley <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>>:<br>
>> >> > On Wed, Jun 10, 2015 at 9:42 AM, Tim Steinhoff<br>
>> >> > <<a href="mailto:kandanovian@gmail.com">kandanovian@gmail.com</a>><br>
>> >> > wrote:<br>
>> >> >><br>
>> >> >> Hi all<br>
>> >> >><br>
>> >> >> I want to use Petsc to solve some linear systems via the built-in<br>
>> >> >> Krylov<br>
>> >> >> subspace methods as well as by means of UMFPACK.<br>
>> >> >><br>
>> >> >> The considered matrix is block sparse with blocks of size 6x6.<br>
>> >> >><br>
>> >> >> Here is what I came up with after taking a look at some of the<br>
>> >> >> examples<br>
>> >> >><br>
>> >> >> MPI_Comm comm;<br>
>> >> >> Mat A;<br>
>> >> >> PetscInt n = 10000; /* dimension of matrix */<br>
>> >> >> comm = PETSC_COMM_SELF;<br>
>> >> >> MatCreate(comm,&A);<br>
>> >> >> MatSetSizes(Amat,n,n,n,n);<br>
>> >> >> MatSetBlockSize(A,6);<br>
>> >> >> MatSetType(A,MATAIJ); /* UMFPACK compatible format due to comm =<br>
>> >> >> PETSC_COMM_SELF */<br>
>> >> >><br>
>> >> >> Questions:<br>
>> >> >> 1.<br>
>> >> >> I work on a single node with 2-8 cores. Hence, comm =<br>
>> >> >> PETSC_COMM_SELF;<br>
>> >> >> I<br>
>> >> >> guess. Is it correct in this contect to set<br>
>> >> >> MatSetSizes(Amat,n,n,n,n);<br>
>> >> >> with<br>
>> >> >> 4-times n?<br>
>> >> ><br>
>> >> ><br>
>> >> > Yes.<br>
>> >> ><br>
>> >> >><br>
>> >> >> 2.<br>
>> >> >> After the above sequence of commands do I have to use something like<br>
>> >> >> MatSeqAIJSetPreallocation(A,0,d_nnz); /* d_nnz <-> number of<br>
>> >> >> nonzeros<br>
>> >> >> per row */<br>
>> >> >> or is it possible to use<br>
>> >> >> MatSeqBAIJSetPreallocation(A,6,0,db_nnz); /* db_nnz <-> number of<br>
>> >> >> block<br>
>> >> >> nonzeros per block row */<br>
>> >> ><br>
>> >> ><br>
>> >> > You should use this if using MATBAIJ.<br>
>> >> ><br>
>> >> >><br>
>> >> >> In any case, is something like<br>
>> >> >> MatSetValuesBlocked(A,1,idx_r,1,idx_c,myblockvals,INSERT_VALUES);<br>
>> >> >> to fill values of one block into the matrix A ok?<br>
>> >> ><br>
>> >> ><br>
>> >> > Yes, however in order to get improved performance, you need type<br>
>> >> > MATBAIJ.<br>
>> >> ><br>
>> >> > Thanks,<br>
>> >> ><br>
>> >> > Matt<br>
>> >> ><br>
>> >> >><br>
>> >> >> Regards<br>
>> >> >> Tim<br>
>> >> ><br>
>> >> ><br>
>> >> ><br>
>> >> ><br>
>> >> > --<br>
>> >> > What most experimenters take for granted before they begin their<br>
>> >> > experiments<br>
>> >> > is infinitely more interesting than any results to which their<br>
>> >> > experiments<br>
>> >> > lead.<br>
>> >> > -- Norbert Wiener<br>
>> ><br>
>> ><br>
>> ><br>
>> ><br>
>> > --<br>
>> > What most experimenters take for granted before they begin their<br>
>> > experiments<br>
>> > is infinitely more interesting than any results to which their<br>
>> > experiments<br>
>> > lead.<br>
>> > -- Norbert Wiener<br>
><br>
><br>
><br>
<span class="HOEnZb"><font color="#888888">><br>
> --<br>
> What most experimenters take for granted before they begin their experiments<br>
> is infinitely more interesting than any results to which their experiments<br>
> lead.<br>
> -- Norbert Wiener<br>
</font></span></blockquote></div><br><br clear="all"><div><br></div>-- <br><div class="gmail_signature">What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>-- Norbert Wiener</div>
</div></div>