[MPICH] MPICH2 status on heterogeneousness support

[William Gropp]

Yes, that's what we're planning on.  MPICH1 could use XDR, which did  
allow for support of non IEEE systems (still plentiful when MPICH1  
was started).  For the first cut, we're planning on only 32/64 and  
different MSB/LSB orderings, though the implementation will make it  
relatively easy to add others as necessary.

Bill

On Oct 19, 2006, at 9:14 AM, Matthew Chambers wrote:

> I hope there are plans to at least support differing type sizes (32  
> bit, 64
> bit, possibly 16 bit interoperability) and differing byte orders  
> (little
> endian and big endian).  Did MPICH1 support different floating point
> representations (other than IEEE)?  The 16 bit support and the  
> different
> floating point representations are insubstantial issues I think,  
> and I'm not
> aware of any other types of heterogeneousness in workstation networks.
>
> -Matt Chambers
>
>> -----Original Message-----
>> From: owner-mpich-discuss at mcs.anl.gov [mailto:owner-mpich-
>> discuss at mcs.anl.gov] On Behalf Of Irfan Syed
>> Sent: Wednesday, October 18, 2006 7:14 PM
>> To: mpich-discuss at mcs.anl.gov
>> Subject: Re: [MPICH] MPICH2 status on heterogeneousness support
>>
>> Hello,
>>
>> What is the scope of 'heterogeneity' which mpich2 plans to support?
>>
>> Thanks,
>> Irfan
>>
>> Rajeev Thakur wrote:
>>> It will probably not be there in the next release, which should  
>>> be in
>>> a month or so. The following release may have some preliminary  
>>> support
>>> for heterogeneous.
>>> Rajeev
>>>
>>>      
>>> --------------------------------------------------------------------
>> ----
>>>     *From:* owner-mpich-discuss at mcs.anl.gov
>>>     [mailto:owner-mpich-discuss at mcs.anl.gov] *On Behalf Of *Matthew
>>>     Chambers
>>>     *Sent:* Tuesday, October 17, 2006 4:44 PM
>>>     *To:* mpich-discuss at mcs.anl.gov
>>>     *Subject:* [MPICH] MPICH2 status on heterogeneousness support
>>>
>>>     I'm curious how close MPICH2 is to supporting heterogeneous
>>>     networks of workstations.
>>>
>>>     Thanks,
>>>
>>>     Matt Chambers
>>>
>
>