[petsc-dev] usage of Malloc2 at VecScatterCopy_MPI_ToAll

[Lisandro Dalcin]

On 9 July 2010 22:07, Barry Smith <bsmith at mcs.anl.gov> wrote:
>
>  Hmm, looks like we could use a Malloc3() there :-) Even better.
>

Is this safe regarding memory alignment?

>
>   Barry
>
> On Jul 9, 2010, at 8:00 PM, Lisandro Dalcin wrote:
>
>> On 9 July 2010 21:26, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>>
>>>  I consider this bogus warnings.
>>>
>>
>> I agree on that. My point is the actual usage of Malloc2.
>>
>>  ierr                =
>> PetscMalloc2(1,VecScatter_MPI_ToAll,&sto,size,PetscMPIInt,&sto->count);CHKERRQ(ierr);
>>  ierr                =
>> PetscMalloc(size*sizeof(PetscMPIInt),&sto->displx);CHKERRQ(ierr);
>>
>> What's the point of allocating 1 struct + size ints, and next size
>> more ints? I would use Malloc for the struct, and then Malloc2 for the
>> two arrays.
>>
>>
>>>   Barry
>>>
>>> On Jul 9, 2010, at 7:22 PM, Lisandro Dalcin wrote:
>>>
>>>> While building with nvcc, I looked at this warning:
>>>>
>>>> libfast in: /usr/local/petsc/dev/src/vec/vec/utils
>>>> vscat.c(322): warning: variable "sto" is used before its value is set
>>>>
>>>> Do we really need to use Malloc2 in such line? Perhaps we should
>>>> actually use Malloc2 for sto->count and sto->displx ?
>>>>
>>>> --
>>>> 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
>
>



-- 
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