[mpich2-dev] [opa-core] Problem with MPICH2 1.1 during MPI_Init() on PPC Linux 32bit

[Darius Buntinas]

The assert is there to make sure that the relative cell pointer is the
same size as the absolute cell pointer, because we want to cast between
the two, so I think the assert is correct.  I'll fix the comment to be
more accurate.

Thanks for checking this out for us!

-d

On 06/09/2009 04:28 PM, Joe Ratterman wrote:
> Yes, that patch works fine.  It might not be the final solution, but it does
> make the assert pass, and I can run my app.  I think it is a very good idea.
> What about this part of the comment:
> 
> "Also, we don't support compilers which add random padding to the start of
> the structure."
> 
> Essentially, the compiler added 4 bytes to the *end* of a different
> structure that you are using in your new structure.  Might that be a
> concern?
> 
> 
> On Tue, Jun 9, 2009 at 4:06 PM, Darius Buntinas <buntinas at mcs.anl.gov>wrote:
> 
>> Joe, can you check if this patch works for you?
>>
>> Thanks,
>> -d
>>
>> Index: src/mpid/ch3/channels/nemesis/nemesis/src/mpid_nem_init.c
>> ===================================================================
>> --- src/mpid/ch3/channels/nemesis/nemesis/src/mpid_nem_init.c   (revision
>> 4676)
>> +++ src/mpid/ch3/channels/nemesis/nemesis/src/mpid_nem_init.c   (working
>> copy)
>> @@ -86,7 +86,7 @@
>>       * should only contain a pointer. Also, we don't support compilers
>>      * which add random padding to the start of the structure. The
>>      * below assert should blow up if any of these happens. */
>> -    MPIU_Assert(sizeof(MPID_nem_cell_rel_ptr_t) == SIZEOF_VOID_P);
>> +    MPIU_Assert(sizeof(MPID_nem_cell_rel_ptr_t) == sizeof(OPA_ptr_t));
>>
>>     /* Make sure the cell structure looks like it should */
>>     MPIU_Assert(MPID_NEM_CELL_PAYLOAD_LEN + MPID_NEM_CELL_HEAD_LEN ==
>> sizeof(MPID_nem_cell_t));
>>
>>
>>
>>
>> On 06/09/2009 03:58 PM, Darius Buntinas wrote:
>>> I believe it should be exactly the size of an OPA_ptr_t.  I'm checking
>>> this right now.
>>>
>>> -d
>>>
>>>> I think the bigger question is whether or not the assertion is
>>>> sensible.  I don't know why the cell rel pointer structure needs to be
>>>> exactly the same size as a pointer as long as it contains a pointer
>>>> value...
>> _______________________________________________
>> opa-core mailing list
>> opa-core at lists.mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/opa-core
>>
>