[petsc-users] Filling a matrix by chunks of nonzero entries

[marco restelli]

2016-06-20 18:00 GMT+0200, Matthew Knepley <knepley at gmail.com>:
> On Mon, Jun 20, 2016 at 8:13 AM, marco restelli <mrestelli at gmail.com>
> wrote:
>
>> Dear all,
>>    while assembling a matrix in PETsc we have the following pattern:
>>
>>
>> do i=1,number_of_chunks
>>
>>  ! generate a chunk of matrix entries
>>  call compute_ith_chunk( i_idx , j_idx , values )
>>
>>  ! insert those entries
>>  do j=1,size_of_the_ith_chunk
>>    call MatSetValue( mat , i_idx(j),j_idx(j),value(j) , add_values )
>>  enddo
>>
>> enddo
>>
>>
>> The problem is that inserting the elements with MatSetValue seems to
>> have a significant overhead due to memory allocations and
>> deallocations.
>>
>> Is there a way to speed-up this process, preallocating memory?
>>
>> Notice that we know the number of elements that we have to insert for
>> each chunk, but we don't know to which extent they overlap, i.e. we do
>> not know how many nonzero entries will result in the final matrix.
>>
>> Also, the entries do not have a block pattern, so we can not use
>> MatSetValues.
>>
>
> Run the whole process once to count the entries, and then insert.

Matt, when you say "insert" do you mean using MatSetValue? That is
what we are currently doing, and it seems to have a significant
overhead due to memory allocations and deallocations.

I know already the number of entries that I am going to insert, what I
don't know is the number of nonzeros in the resulting matrix, since my
entries might or might not overlap. Counting the number of nonzeros
seems to me a nontrivial task, which I would be happy to leave to
PETsc.

Marco