[mvapich-discuss] Building MVAPICH2 1.8a2/1.7 difficulties with PGI 12.2

[Jonathan Perkins]

On Mon, Feb 27, 2012 at 10:27:21AM -0500, Thompson, Matt (GSFC-610.1)[SCIENCE SYSTEMS AND APPLICATIONS INC] wrote:
> On 02/27/2012 10:01 AM, Jonathan Perkins wrote:
> > On Fri, Feb 24, 2012 at 02:10:36PM -0600, Thomas Rothrock CTR SMDC SimCtr/GaN Corporation wrote:
> >> This is a problem in PGI 12.2-0 and affects more than just the MVAPICH2 1.8 trunk.  My first workaround was the following patch:
> >>
> >> I received the following response from PGI earlier today:
> >>
> >>         Unfortunately, the correction did not make it into the 12.2 release
> >>
> >>         In the two files listed below,
> >>
> >>         change
> >>         #if !defined(_GNU_SOURCE)
> >>         to
> >>         #if !defined(_GNU_SOURCE) || !defined(__cplusplus)
> >>
> >>         /opt/pgi/linux86-64/12.2/include/float.h
> >>         /opt/pgi/linux86-64/12.2/include/limits.h
> >>
> >> The latter is likely the best workaround, but relies on you being able to modify the PGI 12.2 installation on affected systems.
> > 
> > Tom, thanks for updating the list with your work around and PGI's
> > response.
> > 
> > Matt, can you verify whether this works for you or not?  Thanks in
> > advance.
> 
> Yes indeed, changing float.h in this way allowed us to build MVAPICH2
> 1.8a2 with PGI 12.2. I'm now doing builds and tests to make sure
> MVAPICH2 built correctly, but it looks good.
> 
> (NOTE: PGI did update the forum topic I started about this by saying the
> best thing to do might be to remove the corresponding #ifdefs altogether
> from limits.h and float.h. This is mainly because Tom, being faster than
> I, tried to make Open MPI and ran into an issue as well concerning these.)
> 
> Thanks for all the help with this,
> Matt

Thanks for the confirmation.  We'll add this to our FAQ.

-- 
Jonathan Perkins
http://www.cse.ohio-state.edu/~perkinjo