[MITgcm-support] Error reading dx file with pathscale

[Martin Losch]

David,

you need to specify either one, but not both together, that would be "double byte swap".

Martin

----- Original Message -----
From: David Hebert <david.hebert.ctr at nrlssc.navy.mil>
Date: Wednesday, October 22, 2008 16:43
Subject: Re: [MITgcm-support] Error reading dx file with pathscale
To: mitgcm-support at mitgcm.org

> 
>

> 
> 
> 
>

> Martin,
> 
> 
> 
> If I specify -D_BYTESWAPIO do I need to specify -convert big_endian in
> compiler? Or will this "double byte swap?"
> 
> 
> 
> Oh, I tried specifying delX and delY as you suggested. Those are
> correct in STDOUT when simulation runs. I then get an error that CG2D
> does not converge and from STDOUT the boundary values look like there
> is an endian issue. 
> 
> 
> 
> Thanks,
> 
> 
> 
> David
> 
> 
> 
> Martin Losch wrote:
>

David,
> 
> 
> 
> 
> good guess, I forgot to mention the wordlength. Sometimes that's a
> problem, too, but you'd have to consult the man-pages or documentation
> for that.
> 
> 
> 
> 
> Did you try the -D_BYTESWAPIO option, just to rule out this
> byteswapping issue?
> 
> 
> 
> 
> Does the model read other fields correctly (try by specifying a
> delX/delY instead of files for this test)?
> 
> 
> 
> 
> Martin
> 
> 
> 
> 
> On 21 Oct 2008, at 20:24, David Hebert wrote:
> 
> 
> 
> 
>
Hi Martin,
> 
> 
> 
> 
> Thanks for responding. Pathscale has both a -convert big_endian and a
> -byteswapio flag which do the same thing. I have tried both options
> with the same result.
> 
> 
> 
> 
> I note that WORDLENGTH=4 for most all compilers. I wonder if there
> would be a reason where I would need to change this for pathscale on
> 64bit architecture?
> 
> 
> 
> 
> 
> 
> 
> 
> David
> 
> 
> 
> 
> 
> 
> Martin Losch wrote:
> 
> 
>

> 
> Hi David,
> 
> 
> just to make sure: you need either set a CPP flag like this:
> 
> 
> DEFINES='-D_BYTESWAPIO'
> 
> 
> in the build options file or (which is probably faster) specify the
> appropiate compiler flag.
> 
> 
> "-byteswapio" is the flag for the PGI compiler, I don't know what the
> Pathscale compiler does, but in the
> tools/build_options/linux_amd64_pathf90 I found this:
> 
> 
> FFLAGS='-convert big_endian'
> 
> 
> 
> 
> Could that be the problem?
> 
> 
> 
> 
> Martin
> 
> 
> On 21 Oct 2008, at 19:37, David Hebert wrote:
> 
> 
> 
> 
>
Hello all,
> 
> 
> 
> 
> 
> 
> I am working on a Cray XT4 system. I was able to compile and run with
> PGI compilers. I am not attempting to see if Pathscale compilers will
> make code run faster. In data file I specify a delXfile and delYfile.
> It seems that by switching compilers now causes the code to read these
> files incorrectly, as I am now getting values like 9.0e160. This seems
> to suggest that byte swapping is not occurring, but I have checked many
> times and I do have -byteswapio flag set at compile time. Any ideas
> what might be causing this behavior?
> 
> 
> 
> 
> Thanks,
> 
> 
> 
> 
> David
> 
> 
> 
> 
> 
> 
> _______________________________________________
> 
> 
> MITgcm-support mailing list
> 
> 
> MITgcm-support at mitgcm.org
> 
> 
> http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
>

> 
> 
> _______________________________________________
> 
> 
> MITgcm-support mailing list
> 
> 
> MITgcm-support at mitgcm.org
> 
> 
> http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
>

> 
> 
> -- 
> 
> David A. Hebert
> 
> 
> NRC Postdoc
> 
> 
> Naval Research Lab
> 
> 
> Stennis Space Center, MS 39525
> 
> 
> david.hebert.ctr at nrlssc.navy.mil
> 
> 
> Phone: (228) 688-5846
> 
> 
> Fax:     (228) 588-5997
> 
> 
> _______________________________________________
> 
> 
> MITgcm-support mailing list
> 
> 
> MITgcm-support at mitgcm.org
> 
> 
> http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
>

> 
> 
> _______________________________________________
> 
> 
> MITgcm-support mailing list
> 
> 
> MITgcm-support at mitgcm.org
> 
> 
> http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
>

> 
> 
>
-- 
> 
> 
>

>
David A. Hebert
> 
> NRC Postdoc
> 
> Naval Research Lab
> 
> Stennis Space Center, MS 39525

>
david.hebert.ctr at nrlssc.navy.mil
> 
> Phone: (228) 688-5846
> 
> Fax:     (228) 588-5997
> 
>

>

> 
> 
> 
> 
> 
>
>

> 
> > _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support

Martin Losch
Alfred Wegener Institute 
Postfach 120161, 27515 Bremerhaven, Germany; 
Tel./Fax: ++49(0471)4831-1872/1797




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20081022/3ea5fa06/attachment.htm>