[mpich-discuss] Segmentation fault when MPICH is compiled with Portland compiler 17.5

[Jeff Hammond]

I reproduced this with PGI 17.4 community edition on Linux with your
Fortran test and an equivalent C one.

GDB indicates the SEGV happens here:

Program received signal SIGSEGV, Segmentation fault.
MPID_nem_init (pg_rank=0, pg_p=0x60fce0, has_parent=8388607) at
../src/mpid/ch3/channels/nemesis/src/mpid_nem_init.c:305
305 MPID_nem_queue_enqueue(MPID_nem_mem_region.FreeQ[pg_rank],
&(MPID_nem_mem_region.Elements[idx]));

Jeff

On Tue, Jul 11, 2017 at 1:19 AM, Andrew Wood <andrew at fluidgravity.co.uk>
wrote:
>
> > It's hard not to conclude this is a PGI 17 bug from your description.
Have you
> > reported this to PGI? I would hate for the MPICH team to take support
work away
> > from the people PGI pays to do this.
>
> Indeed, and I have done so.
>
> However, in my experience, it tends to be somewhat unpredictable whether
> incorrect code causes a segmentation fault or not. So I thought it worth
> checking here too.
>
> Regards,
> Andy.
>
> On 10/07/17 21:17, Jeff Hammond wrote:
> > "This occurs with PGI versions 17.1 and 17.5, but not with versions
16.1, 16.5 or
> > 16.7, nor with the current Intel or gcc compilers."
> >
> > It's hard not to conclude this is a PGI 17 bug from your description.
Have you
> > reported this to PGI? I would hate for the MPICH team to take support
work away
> > from the people PGI pays to do this.
> >
> > Jeff
> >
> >
> > On Mon, Jul 10, 2017 at 6:48 AM Andrew Wood <andrew at fluidgravity.co.uk
> > <mailto:andrew at fluidgravity.co.uk>> wrote:
> >
> >     Hi,
> >
> >     I'm getting a segmentation fault in MPI_INIT, when MPICH is
compiled with the
> >     current version of the Portland compilers:
> >
> >     > mpif90 code.f90 && ./a.out
> >     Segmentation fault
> >
> >     code.f90 is in the attached .tar.gz file, along with the .log files
generated by
> >     configuring and building MPICH.
> >
> >     This occurs with PGI versions 17.1 and 17.5, but not with versions
16.1, 16.5 or
> >     16.7, nor with the current Intel or gcc compilers. It occurs with
MPICH versions
> >     3.1.4, 3.2 and the current master branch in the github repository.
> >
> >     The problem does not occur if MPICH is configured with the option
> >     '--disable-fast'.
> >
> >     Is this a bug in MPICH or PGI, or have I done something else wrong?
> >
> >     Thanks,
> >     Andy.
> >
> >     --
> >     Dr Andrew Wood
> >     Fluid Gravity Engineering Ltd.
> >     83 Market Street
> >     St Andrews
> >     Fife KY16 9NX
> >     Tel: +44 (0)1334 460805
> >     Fax: +44 (0)1334 460813
> >
> >     Fluid Gravity Engineering Ltd is registered in the UK with
registration number
> >     1674369. The registered address is Fluid Gravity Engineering Ltd,
Unit 1, The
> >     Old Coach House, 1 West Street, Emsworth, Hampshire, PO10 7DX.
> >     _______________________________________________
> >     discuss mailing list     discuss at mpich.org <mailto:discuss at mpich.org
>
> >     To manage subscription options or unsubscribe:
> >     https://lists.mpich.org/mailman/listinfo/discuss
> >
> > --
> > Jeff Hammond
> > jeff.science at gmail.com <mailto:jeff.science at gmail.com>
> > http://jeffhammond.github.io/
>
>
> --
> Dr Andrew Wood
> Fluid Gravity Engineering Ltd.
> 83 Market Street
> St Andrews
> Fife KY16 9NX
> Tel: +44 (0)1334 460805
> Fax: +44 (0)1334 460813
>
> Fluid Gravity Engineering Ltd is registered in the UK with registration
number
> 1674369. The registered address is Fluid Gravity Engineering Ltd, Unit 1,
The
> Old Coach House, 1 West Street, Emsworth, Hampshire, PO10 7DX.
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss




--
Jeff Hammond
jeff.science at gmail.com
http://jeffhammond.github.io/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20170711/76fe28bc/attachment.html>
-------------- next part --------------
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss