[petsc-users] Direct solvers

[Barry Smith]

  Is this still an issue for you?  If so could you build both 3.5.1 and 3.5.2 from the tarballs using the --download-xxx options for installing external packages then compile and run the same application with both with the same options and send the output from running both with -log_summary. This will help us see if the the memory/performance between the two has somehow unexpectedly changed.

  Barry

> On Feb 6, 2015, at 8:53 AM, Manav Bhatia <bhatiamanav at gmail.com> wrote:
> 
> 
>> On Feb 5, 2015, at 11:18 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>> 
>>>   I am trying to use an lu decomposition method for a relatively large matrix (~775,000 dofs) coming from a thermoelasticity problem. 
>>> 
>>>   For the past few weeks, LU solver in 3.5.1 has been solving it just fine. I just upgraded to 3.5.2 from macports (running on Mac OS 10.10.2), and am getting the following “out of memory" error 
>> 
>>   This is surprising. The changes are 3.5.1 to 3.5.2 are supposed to be only minor bug fixes. Is the code otherwise __exactly__ the same with the same options?  Are all the external libraries exactly the same in both cases?
> 
> The 3.5.1 version I had been using was one that I had downloaded from the petsc website and compiled myself, without some external packages, like MUMPS, suitesparse and superlu. 
> 
> The 3.5.2 version was build by macports with the added options of MUMPS, suitesparse and superlu. 
> 
> My source code remained the same between these separate runs. 
> 
> I will try rebuilding this from scratch to see if it makes any change. 
> 
> -Manav