On 3 October 2010 19:51, Barry Smith <bsmith at mcs.anl.gov> wrote:
>
> ? Well it is always "with array" since you always need to provide an array of
> indices, so why both with the redundant "withArray" (With
> VecCreateSeqWithArray() there is also a VecCreateSeq() that does not take an
> array, hence WithArray in that case is not redundant)?
>
I'm just thinking on user convenience...
>
> ? Regarding having two routines: I found having three almost identical
> routines unneededly complicated.
>
I definitely agree.
>
> You still want to have two routines, one is just a subset of the
> functionality of the second. That is redundant and hence unneeded.
>
If we were using C++, don't you think we could overload
ISCreateGeneral? But we are in C, and I see that
ISCreateGeneralWithArray() as a developer-level call, while
ISCreateGeneral() looks more like an user-level-call.
I'm proposing making ISCreateGeneral() a wrapper for
ISCreateGeneralWithArray(). I do not think that the redundancy is too
much of a trouble in such cases.
> ? I do appreciate your suggestions, please keep on having them even if they
> don't always convince me.
>
Well, I'm still trying to convince you :-)
>
> ? Barry
>
> On Oct 3, 2010, at 5:45 PM, Lisandro Dalcin wrote:
>
>> On 3 October 2010 14:57, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>>
>>> ? I find the three routines ISCreateGeneral(), ISCreateGeneralNC() and
>>> ISCreateGeneralWithArray() to be confusing and code redundant.
>>>
>>> ? Therefore in PETSc-dev I have introduced
>>>
>>> /*E
>>> ? ?PetscCopyMode ?- Determines how an array passed to certain functions is
>>> copied or retained
>>>
>>> ? Level: beginner
>>>
>>> $ ? PETSC_COPY_VALUES - the array values are copied into new space, the
>>> user is free to reuse or delete the passed in array
>>> $ ? PETSC_OWN_POINTER - the array values are NOT copied, the object takes
>>> ownership of the array and will free it later, the user cannot change or
>>> $ ? ? ? ? ? ? ? ? ? ? ? delete the array. The array MUST have been obtained
>>> with PetscMalloc(). Hence this mode cannot be used in Fortran.
>>> $ ? PETSC_USE_POINTER - the array values are NOT copied, the object uses
>>> the array but does NOT take ownership of the array. The user cannot use
>>> ? ? ? ? ? ? ? ? ? ? ? ?the array but the user must delete the array after
>>> the object is destroyed.
>>>
>>> E*/
>>> typedef enum { PETSC_COPY_VALUES, PETSC_OWN_POINTER, PETSC_USE_POINTER}
>>> PetscCopyMode;
>>> extern const char *PetscCopyModes[];
>>>
>>> and merged ?ISCreateGeneral/NC/WithArray() into a single routine with an
>>> additional PetscCopyMode argument.
>>>
>>> I have also merged ISLocalToGlobalMappingCreate() and
>>> ISLocalToGlobalMappingCreateNC() into a single routine.
>>>
>>> There may be other places we can use this paradigm in the future, keep your
>>> eyes open,
>>>
>>> ?Happy computing,
>>>
>>> ? ?Barry
>>>
>>
>> I would prefer to have:
>>
>> ISCreateGeneralWithArray(comm,n,ptr,copy_mode,&is);
>>
>> and then implement
>>
>> ISCreateGeneral(comm,n,ptr,&is) {
>> return ISCreateGeneralWithArray(comm,n,ptr,PETSC_COPY_VALUES,&is);
>> }
>>
>> What do you think?
>>
>> --
>> Lisandro Dalcin
>> ---------------
>> CIMEC (INTEC/CONICET-UNL)
>> Predio CONICET-Santa Fe
>> Colectora RN 168 Km 472, Paraje El Pozo
>> Tel: +54-342-4511594 (ext 1011)
>> Tel/Fax: +54-342-4511169
>
>
--
Lisandro Dalcin
---------------
CIMEC (INTEC/CONICET-UNL)
Predio CONICET-Santa Fe
Colectora RN 168 Km 472, Paraje El Pozo
Tel: +54-342-4511594 (ext 1011)
Tel/Fax: +54-342-4511169
