[mpiwg-hybridpm] Changes to the EI chapter

Jeff Hammond jeff.science at gmail.com
Tue Mar 3 17:56:09 CST 2015

On Tue, Mar 3, 2015 at 3:48 PM, Jeff Squyres (jsquyres)
<jsquyres at cisco.com> wrote:
> On Mar 3, 2015, at 3:46 PM, Jeff Hammond <jeff.science at gmail.com> wrote:
>>> Not linking against pthreads doesn't mean anything -- every process still has at least 1 thread.
>> You are making the semantic link between an OS process and a thread in
>> CS lexicon.  That's really not helpful here.
> It's a fact: every process, even every MPI process, has at least 1 thread.
> How is that wrong?

It is a logical tautology.  It is not a useful contribution to this discussion.

>>> That was the whole point of the ticket: these functions are *always* thread safe.
>>> ...which is what Jeff Hammond wanted.  :-)
>> Then we are requiring ALL implementations to support threads.  You
>> want to make that rather huge change to MPI as part of dickering over
>> this sentence?
> Ok, I'm lost.  You wanted this ticket.  You started this ticket.  You pushed for this ticket.

Yes, and I expect that an implementation that supports threads in the
most basic sense of the word, which means that it is theoretically
possible that an implementation can return provided>MPI_THREAD_SINGLE,
that it be possible to determine what provided is without causing the
program to be invalid.

If an implementation is incapable by design and documentation to
return provided>MPI_THREAD_SINGLE because it has taken the lame path
of not supporting user threads beyond those that run "int main()",
then it is impossible for us to ask them to support this feature.

> Now you don't want it any more?

See above.

Jeff Hammond
jeff.science at gmail.com

More information about the mpiwg-hybridpm mailing list