libMesh: A C++ Finite Element Library

On Fri, 4 Dec 2009 13:59:46 -0600 (CST), Roy Stogner <roystgnr@...> wrote:
> In the LibMeshInit constructor, we test MPI_Initialized() before
> calling MPI_Init ourselves, so that if we're incorporated into a
> larger MPI-using code we won't step on their toes.
> 
> We don't currently do the same for PETSc or SLEPc.

PetscInitialize (and SlepcInitialize) do nothing if PETSc is already
initialized so that's not a problem, the reason you have to check
PetscInitialized is that you need to know whether you are responsible
for calling PetscFinalize which can't be called multiple times.
(Perhaps there is an argument for making it do nothing if already
called, but doing so is a bug more often than not.)

> I'm not sure what to do about SLEPc - there doesn't appear to be a
> SlepcInitialized function.  Is there some other way to test if
> SlepcInitialize has been called?  Is it safe to call SlepcInitialize
> twice?  Or do I need to pester slepc-maint with a feature request?

extern PetscTruth SlepcInitializeCalled;


Jed

On Fri, Dec 4, 2009 at 1:59 PM, Roy Stogner <roystgnr@...> wrote:
>
> In the LibMeshInit constructor, we test MPI_Initialized() before
> calling MPI_Init ourselves, so that if we're incorporated into a
> larger MPI-using code we won't step on their toes.
>
> We don't currently do the same for PETSc or SLEPc.
>
> I'm going to wrap the PetscInitialize call in a similar
> PetscInitialized test, so that we only initialize PETSc if nobody else
> already has.  This shouldn't break anyone's code, right?
>
> I'm not sure what to do about SLEPc - there doesn't appear to be a
> SlepcInitialized function.  Is there some other way to test if
> SlepcInitialize has been called?  Is it safe to call SlepcInitialize
> twice?  Or do I need to pester slepc-maint with a feature request?

In the latest slepc (3.0.0-p7) there is an

extern PetscTruth SlepcInitializeCalled;

in include/slepc.h.  Maybe we could check the value of that?

-- 
John

On Fri, 4 Dec 2009, John Peterson wrote:

> On Fri, Dec 4, 2009 at 1:59 PM, Roy Stogner <roystgnr@...> wrote:
>>
>> In the LibMeshInit constructor, we test MPI_Initialized() before
>> calling MPI_Init ourselves, so that if we're incorporated into a
>> larger MPI-using code we won't step on their toes.
>>
>> We don't currently do the same for PETSc or SLEPc.
>>
>> I'm going to wrap the PetscInitialize call in a similar
>> PetscInitialized test, so that we only initialize PETSc if nobody else
>> already has.  This shouldn't break anyone's code, right?
>>
>> I'm not sure what to do about SLEPc - there doesn't appear to be a
>> SlepcInitialized function.  Is there some other way to test if
>> SlepcInitialize has been called?  Is it safe to call SlepcInitialize
>> twice?  Or do I need to pester slepc-maint with a feature request?
>
> In the latest slepc (3.0.0-p7) there is an
>
> extern PetscTruth SlepcInitializeCalled;
>
> in include/slepc.h.  Maybe we could check the value of that?

Sounds good.  It looks like SlepcInitializeCalled was added in their
r1409, which got it first into 2.3.3-p1, so I'll test
SLEPC_VERSION_LESS_THAN(2,3,3) before using it.

And thanks very much to Jed for reminding me about the problems of
multiple finalizes; I'll add libmesh_initialized_petsc and
libmesh_initialized_slepc variables for LibMeshInit to track.
---
Roy