[CIG-SEISMO] Some problems with xspecfem3D and the Intel compilers

[Dimitri Komatitsch]

Hi Rama, Hi all,

Thanks a lot. That would be very useful. It seems option -assume 
buffered_io has a problem in ifort v17, but it would be a pity to get 
rid of it because when it works it does speed up the code significantly.

I attach another bug we have detected with v17 (maybe the same bug?), 
see the comment at line 494 in the attached file (that file is 
src/shared/read_value_parameters.f90 in the Git version), which is from 
SPECFEM2D (to get it: git clone --recursive --branch devel 
https://github.com/geodynamics/specfem2d.git )

Thank you very much,
Best regards,
Dimitri.

On 02/10/2018 06:50 AM, Malladi, Rama Kishan V wrote:
> Hi Dimitri, All,
> I will check what changed in the recent version of the Intel compiler and see if this can be reported to the compiler team as a bug report.
> 
> Thanks
> -Rama
> 
> -----Original Message-----
> From: CIG-SEISMO [mailto:cig-seismo-bounces at geodynamics.org] On Behalf Of Dimitri Komatitsch
> Sent: Friday, February 9, 2018 6:55 PM
> To: cig-seismo at geodynamics.org; Fabien Dubuffet <fabien.dubuffet at univ-lyon1.fr>; Larmat Carene <carene at lanl.gov>
> Subject: Re: [CIG-SEISMO] Some problems with xspecfem3D and the Intel compilers
> 
> 
> Hi Carène and Fabien, Hi all,
> 
> Thanks a lot! Since several users have reported problems with this Intel compiler option, in particular with ifort version 17, let me keep it but turn it off by default (in 2D, 3D, and 3D_GLOBE).
> 
> When it works fine it does speed up I/Os significantly though, thus worth trying it. I will mention that in the users manual (of 2D, 3D, and 3D_GLOBE).
> 
> Thanks!
> Best regards,
> Dimitri.
> 
> On 02/09/2018 11:04 AM, Fabien Dubuffet wrote:
>> Hello Carene,
>>
>> I have tried your solution, and the '-assume nobuffered_io' flag
>> solves this problem.
>> Thanks a lot,
>>
>> Cheers,
>>
>> Fabien
>>
>> Le 09/02/18 à 00:49, Larmat, Carene a écrit :
>>> I had some success bypassing some intel compiler issues using -assume
>>> nobuffered_io for those of us who are imposed to work with new
>>> versions.
>>> Cheers!
>>>
>>> Carene Larmat, EES-17, MS D452
>>> carene at lanl.gov <mailto:carene at lanl.gov>, 505 667 2074
>>
>>
>> _______________________________________________
>> CIG-SEISMO mailing list
>> CIG-SEISMO at geodynamics.org
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>
> 
> --
> Dimitri Komatitsch, CNRS Research Director (DR CNRS) Laboratory of Mechanics and Acoustics, Marseille, France http://komatitsch.free.fr _______________________________________________
> CIG-SEISMO mailing list
> CIG-SEISMO at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
> _______________________________________________
> CIG-SEISMO mailing list
> CIG-SEISMO at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
> 

-- 
Dimitri Komatitsch, CNRS Research Director (DR CNRS)
Laboratory of Mechanics and Acoustics, Marseille, France
http://komatitsch.free.fr
-------------- next part --------------
A non-text attachment was scrubbed...
Name: read_value_parameters.f90
Type: text/x-fortran
Size: 13901 bytes
Desc: not available
URL: <http://lists.geodynamics.org/pipermail/cig-seismo/attachments/20180210/8cefdbb2/attachment.bin>