[mpich-discuss] Building MPICH 3.2 on Power8 with PGI

Balaji, Pavan balaji at anl.gov
Wed Mar 23 08:45:03 CDT 2016


Can you try the suggestion in the error message below?

  -- Pavan

> On Mar 23, 2016, at 5:21 AM, Bader, Reinhold <Reinhold.Bader at lrz.de> wrote:
> 
> Dear MPICH developers,
>  
> I am currently trying to build MPICH 3.2 with the PGI 16.3 beta release on little-endian Power8 running Ubuntu.
> I am failing with this in the configure stage. I have tried the following two:
>  
> ./configure --prefix=/opt/mpich_pgi --with-pm=hydra CC=pgcc FC=pgfortran CPP="pgcc -E"
>  
> This fails with a compilation error that can be reproduced with the following program:
>  
> #define int8_t signed char
> #include <sys/types.h>
> int main ()
> {
>   return 0;
> }
>  
> … with an obviously misplaced macro definition for int8_t.
>  
> The second scenario uses
> ./configure --disable-cxx --prefix=/opt/mpich_pgi_nocxx --with-pm=hydra CC=pgcc FC=pgfortran CPP="pgcc -E"
>  
> and fails with the error message
> “checking for OpenPA atomic primitive availability... no
> configure: error: 
> The ch3 device was selected yet no native atomic primitives are
> available on this platform.  OpenPA can emulate atomic primitives using
> locks by specifying --with-atomic-primitives=no but performance will be
> very poor.  This override should only be specified for correctness
> testing purposes.”
>  
> Unfortunately the config.log only shows an exit status of 1, and not any compiler error message or test program
> that contains executable statements.
>  
> Any ideas?
>  
> Regards
>  
>  
> --
> Dr. Reinhold Bader
> Leibniz Supercomputing Centre (http://www.lrz.de) / HPC Systems and Services
> Tel.: 0049 89 35831 8825  Fax: 8625
>  
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss


More information about the discuss mailing list