[Trilinos-Users] run time changing matrix

[Lucia Mirabella]

Thanks Chris and Alan for your suggestions.

Actually I have realized that the cpu time required for the solution of 
the linear system doesn't improve if I eliminate some zero elements from 
the pattern, so, as Alan pointed out, it is not really convenient to 
modify the matrix pattern at each time step. I think I will pad the 
matrix, as Alan suggested.

Thanks,

Lucia


Williams, Alan B wrote:
> Lucia,
> Chris baker points out the following:
>
> Chris Baker <cgbaker at sandia.gov> wrote:
>
> I believe that you can add entries to a matrix after FillComplete(), as long as you do not call OptimizeStorage(). FillComplete() needs to be called after adding entries, before the matrix should be used again. I wrote a simple serial test demonstrating this, output is below.
>
> Removing entries is more difficult. The easiest way that I can think of is to ReplaceMyValues() the entry-to-be-removed with an existing column index (the diagonal, perhaps) and the value zero; it will be eliminated on the call to MergeRedundantEntries() in the next FillComplete(). 
>
> Chris
>
> [s903186:~/EPETRA_SERIAL] cgbaker% ./Lucia_test.exe
>
>
> ****** Created matrix
>
>
> ****** Added entries to main diagonal: 
>
> Number of Global Rows        = 4
> Number of Global Cols        = 4
> Number of Global Diagonals   = 4
> Number of Global Nonzeros    = 4
> Global Maximum Num Entries   = 1
>  ** Matrix is Lower Triangular **
>  ** Matrix is Upper Triangular **
>
>
>
> Number of My Rows        = 4
> Number of My Cols        = 4
> Number of My Diagonals   = 4
> Number of My Nonzeros    = 4
> My Maximum Num Entries   = 1
>
>    Processor    Row Index    Col Index           Value     
>        0             0             0                       0    
>        0             1             1                       0    
>        0             2             2                       0    
>        0             3             3                       0    
>
>
> ****** Added entries to first super-diagonal: 
>
> Number of Global Rows        = 4
> Number of Global Cols        = 4
> Number of Global Diagonals   = 4
> Number of Global Nonzeros    = 7
> Global Maximum Num Entries   = 2
>  ** Matrix is Lower Triangular **
>  ** Matrix is Upper Triangular **
>
>
>
> Number of My Rows        = 4
> Number of My Cols        = 4
> Number of My Diagonals   = 4
> Number of My Nonzeros    = 7
> My Maximum Num Entries   = 2
>
>    Processor    Row Index    Col Index           Value     
>        0             0             0                       0    
>        0             0             1                       1    
>        0             1             1                       0    
>        0             1             2                       1    
>        0             2             2                       0    
>        0             2             3                       1    
>        0             3             3                       0    
>
>
> ****** Added entries to first sub-diagonal: 
>
> Number of Global Rows        = 4
> Number of Global Cols        = 4
> Number of Global Diagonals   = 4
> Number of Global Nonzeros    = 10
> Global Maximum Num Entries   = 3
>  ** Matrix is Lower Triangular **
>  ** Matrix is Upper Triangular **
>
>
>
> Number of My Rows        = 4
> Number of My Cols        = 4
> Number of My Diagonals   = 4
> Number of My Nonzeros    = 10
> My Maximum Num Entries   = 3
>
>    Processor    Row Index    Col Index           Value     
>        0             0             0                       0    
>        0             0             1                       1    
>        0             1             0                      -1    
>        0             1             1                       0    
>        0             1             2                       1    
>        0             2             1                      -1    
>        0             2             2                       0    
>        0             2             3                       1    
>        0             3             2                      -1    
>        0             3             3                       0    
>  
>
>   
>> -----Original Message-----
>> From: trilinos-users-bounces at software.sandia.gov 
>> [mailto:trilinos-users-bounces at software.sandia.gov] On Behalf 
>> Of Williams, Alan B
>> Sent: Monday, May 04, 2009 11:49 AM
>> To: Lucia Mirabella; trilinos-users at software.sandia.gov
>> Subject: Re: [Trilinos-Users] run time changing matrix
>>
>> Lucia,
>>
>> The structure of an Epetra matrix can't change after 
>> FillComplete has been called.
>>
>> We have some applications which require changing the 
>> structure of the matrix, and we destroy and recreate the 
>> matrix when that is necessary.
>>
>> If your structural changes are minor, and the number of rows 
>> stays constant, it may be possible to pad your matrix a 
>> little bit and store zeros for elements that are not in use.
>>
>> Alan
>>
>>
>>     
>>> -----Original Message-----
>>> From: trilinos-users-bounces at software.sandia.gov 
>>> [mailto:trilinos-users-bounces at software.sandia.gov] On Behalf 
>>> Of Lucia Mirabella
>>> Sent: Thursday, April 30, 2009 5:35 PM
>>> To: trilinos-users at software.sandia.gov
>>> Subject: [Trilinos-Users] run time changing matrix
>>>
>>> Hello,
>>>
>>> I need to use in my code an Epetra_FECrsMatrix that changes 
>>> during the 
>>> execution of the code. In particular the pattern of the 
>>> matrix may vary 
>>> at different time steps, increasing or decreasing the number 
>>> of non-zero 
>>> elements. In general this number is quite small.
>>>
>>> Could you please suggest me the most efficient way (in terms 
>>> of memory 
>>> usage and CPU time to solve the linear system) to implement 
>>> this feature?
>>>
>>> Thank you,
>>>
>>> Lucia
>>>
>>> _______________________________________________
>>> Trilinos-Users mailing list
>>> Trilinos-Users at software.sandia.gov
>>> http://software.sandia.gov/mailman/listinfo/trilinos-users
>>>
>>>       
>> _______________________________________________
>> Trilinos-Users mailing list
>> Trilinos-Users at software.sandia.gov
>> http://software.sandia.gov/mailman/listinfo/trilinos-users
>>
>>     
>
>